Notice of Pre-AIA or AIA Status The present application, filed on or after March 16, 2013, is being examined under the first inventor to file provisions of the AIA. Claim Objections Claim 19 is objected to because of the following informality: claim 19 recites The method of claim 18 , however, claim 18 recites A non-transitory computer readable storage medium tangibly embodying a computer readable program code having computer readable instructions that, when executed, causes a computer device to carry out a method . Therefore claim 19 should properly recite “ The non-transitory computer readable storage medium of claim 18, ” . Appropriate correction is required. Claim 20 is objected to because of the following informality: claim 20 recites The method of claim 18 , however, claim 18 recites A non-transitory computer readable storage medium tangibly embodying a computer readable program code having computer readable instructions that, when executed, causes a computer device to carry out a method . Therefore claim 20 should properly recite “ The non-transitory computer readable storage medium of claim 18, ”. Appropriate correction is required. Claim Rejections - 35 USC § 112(b) The following is a quotation of 35 U.S.C. 112(b): (b) CONCLUSION.—The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the inventor or a joint inventor regards as the invention. The following is a quotation of 35 U.S.C. 112 (pre-AIA), second paragraph: The specification shall conclude with one or more claims particularly pointing out and distinctly claiming the subject matter which the applicant regards as his invention. Claims 7 and 1 6 are rejected under 35 U.S.C. 112(b) or 35 U.S.C. 112 (pre-AIA), second paragraph, as being indefinite for failing to particularly point out and distinctly claim the subject matter which the inventor or a joint inventor (or for applications subject to pre-AIA 35 U.S.C. 112, the applicant), regards as the invention. Regarding claim 7 , Claim 7 recites the term “ the action-value function ” , however it is unclear what this refer s to, as no “ action-value function ” is referred to previously within claim 7 , or any parent claims, al though the similar term “ a value function ” is recited within parent claim 6 . The term “ the action-value function ” thus lacks antecedent basis. For examination purposes, the term will be interpreted as reading “ an action-value function ” within claim 7 . Regarding claim 16 , Claim 16 recites the term “ the action-value function ” , however it is unclear what this refer s to, as no “ action-value function ” is referred to previously within claim 16 , or within parent claim 14, al though the similar term “ a value function ” is recited earlier within claim 16 . The term “ the action-value function ” thus lacks antecedent basis. For examination purposes, the term will be interpreted as reading “ an action-value function ” within claim 16 . Claim Rejections - 35 USC § 101 35 U.S.C. 101 reads as follows: Whoever invents or discovers any new and useful process, machine, manufacture, or composition of matter, or any new and useful improvement thereof, may obtain a No therefor, subject to the conditions and requirements of this title. Claims 1-20 are rejected under 35 U.S.C. 101 because the claimed invention is directed to abstract ideas without significantly more. Regarding claim 1, Step 1 - “Is the claim to a process, machine, manufacture or composition of matter?” Yes, the claim is directed towards a process. Step 2A, Prong 1 - “Is the claim directed to a law of nature, a natural phenomenon (product of nature) or an abstract idea?”: The limitation of defining a reward function based on a state and an action as a linear combination of a plurality of component reward functions and a weight for each of the plurality of component reward functions; recites a mathematical formula of a reward function which linearly combines weights for component reward functions, which is a mathematical concept, which is an abstract idea. The limitation of sampling multiple dimensions of the weight for each of the plurality of component reward functions from a continuous distribution between a maximum weight and a minimum weight; recites a mathematical calculation of sampling a statistical distribution, which is a mathematical concept, which is an abstract idea. Step 2A, Prong 2 - “Does the claim recite additional elements that integrate the judicial exception into a practical application?”: The limitation of and training a single policy of the artificial intelligent agent over a continuous goal space including a plurality of parameterized reward functions represented by the continuous distribution of the weight for each of the plurality of component reward functions recites mere instructions to apply training to an artificial intelligen t agent policy, which does not integrate the exception into a practical application, MPEP 2106.05(d) and 2106.05(f). Step 2B - “Does the claim recite additional elements that amount to significantly more than the judicial exception?”: The limitation of and training a single policy of the artificial intelligent agent over a continuous goal space including a plurality of parameterized reward functions represented by the continuous distribution of the weight for each of the plurality of component reward functions recites mere instructions to apply training to an artificial intelligen t agent policy, which is not significantly more than any recited judicial exceptions , MPEP 2106.05(f). Therefore, claim 1 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 2, Claim 2 adds the additional limitation s to claim 1: further comprising improving a performance of the artificial intelligent agent over a segment of the continuous distribution of the weight by providing a skewed distribution of weight, recites mere instructions to apply performance improvement via a skewed weight distribution, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). wherein the training is performed over the skewed distribution of weight for one or more of the plurality of component reward functions recites mere instructions to apply training over a skewed weight distribution, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 2 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 3, Claim 3 adds the additional limitation to claim 2: wherein the skewed distribution of weight is a log-uniform distribution recites further detail on the skewed distribution of weight, without changing that the skewed distribution of weight is merely applied to improve performance, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 3 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 4, Claim 4 adds the additional limitation to claim 1: further comprising sampling the continuous distribution of weights once per training rollout at a beginning of an episode recites a mathematical calculation of sampling a statistical distribution, which is a mathematical concept, which is an abstract idea. Therefore, claim 4 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 5, Claim 5 adds the additional limitations to claim 1: further comprising repeatedly re-sampling the continuous distribution of weights during a training rollout, recites a mathematical calculation of iteratively sampling a statistical distribution, which is a mathematical concept, which is an abstract idea. wherein the artificial intelligent agent becomes robust to reward function changes during ongoing trajectories recites mere instructions to apply increased robustness, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 5 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 6, Claim 6 adds the additional limitation to claim 1: further comprising applying the continuous distribution of weights to both a policy and a value function of a training algorithm recites mere instructions to apply a weight distribution, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 6 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 7, Claim 7 adds the additional limitation to claim 6: further comprising updating a neural network policy from π(s) to π( s,ŵ ) and the action-value function Q(s, a) to Q(s, a, ŵ) by concatenating the continuous distribution of weights, ŵ with inputs related to state, s recites a mathematical formulas for a network policy and for an action-value function, which are mathematical concepts, which are abstract ideas. Therefore, claim 7 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 8, Claim 8 adds the additional limitations to claim 1: further comprising evaluating the single policy of the artificial intelligent agent at inference time by choosing a chosen weight for each of the plurality of component reward functions, recites an evaluation of a policy and a judgement of a weight to choose, which are mental processes, which are abstract ideas, regardless of if they’re performed on a generic computer or using generic machine learning models. wherein the artificial intelligent agent behaves accordingly under a chosen reward function without any retraining recites mere instructions to apply a chosen reward function to an artificial intelligent agent , which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 8 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 9, Claim 9 adds the additional limitation to claim 1: wherein the artificial intelligent agent operates in a racing game environment recites mere instructions to apply the artificial intelligent agent to a racing game environment, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 9 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 10, Claim 10 adds the additional limitations to claim 9: providing a base reward as one of the plurality of component reward functions, the base reward motivating the artificial intelligent agent to finish a race in a minimal time; recites further detail on the component reward functions, without changing that the component reward functions are part of a mathematical formula of a reward function which linearly combines weights for the component reward functions, which is a mathematical concept, which is an abstract idea. a nd providing one or more additional ones of the plurality of component reward functions to provide one or more skill component reward functions and/or one or more personality component reward functions recites further detail on the component reward functions, without changing that the component reward functions are part of a mathematical formula of a reward function which linearly combines weights for the component reward functions, which is a mathematical concept, which is an abstract idea. Therefore, claim 10 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 11, Claim 11 adds the additional limitations to claim 10: wherein the continuous distribution of weights for the one or more additional ones describe an importance of each of the one or more additional ones in relation to a fixed weight for the base reward recites a judgement of the relative importance of weights, which is a mental process, which is an abstract idea, regardless of if it’s performed on a generic computer or using a generic machine learning model. Therefore, claim 11 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 12, Claim 12 adds the additional limitation to claim 10: wherein the one or more additional ones of the plurality of component reward functions include at least one of the following: (a) a penalty for degradation of car tires of the artificial intelligent agent during an environment step; (b) a penalty for a fuel use by the artificial intelligent agent during an environment step; (c) a linear penalty for tire slip ratios and angles; (d) a linear positive reward for tire slip ratio and angle; (e) an edge distance penalty that linearly increases with a proximity of the artificial intelligent agent to an edge of a racing track; (f) a set of reward parts penalizing the artificial intelligent agent for driving in corresponding slices of the racing track defined by a distance to a centerline thereof; (g) a passing reward with independently weighted positive and negative parts for overtaking and being overtaken by other cars, respectively; (h) a penalty on a change in steering angle during an environment step; ( i ) a penalty for colliding with other vehicles; and (j) a penalty for a car of the artificial intelligent agent driving off course recites further detail on the component reward functions, without changing that the component reward functions are part of a mathematical formula of a reward function which linearly combines weights for the component reward functions, which is a mathematical concept, which is an abstract idea. Therefore, claim 12 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 13, Claim 13 adds the additional limitation to claim 12: wherein each of the one or more additional ones of the plurality of component reward functions are defined within the single policy of the artificial intelligent agent recites further detail where the component reward functions are defined, without changing that the component reward functions are part of a mathematical formula of a reward function which linearly combines weights for the component reward functions, which is a mathematical concept, which is an abstract idea. Therefore, claim 1 3 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 1 4 , Step 1 - “Is the claim to a process, machine, manufacture or composition of matter?” Yes, the claim is directed towards a process. Step 2A, Prong 1 - “Is the claim directed to a law of nature, a natural phenomenon (product of nature) or an abstract idea?”: The limitation of defining a reward function based on a state and an action as a linear combination of a plurality of component reward functions and a weight for each of the plurality of component reward functions; recites a mathematical formula of a reward function which linearly combines weights for component reward functions, which is a mathematical concept, which is an abstract idea. The limitation of sampling multiple dimensions of the weight for each of the plurality of component reward functions from a continuous distribution between a maximum weight and a minimum weight; recites a mathematical calculation of sampling a statistical distribution, which is a mathematical concept, which is an abstract idea. The limitation of wherein the plurality of component reward functions include a base reward, motivating the artificial intelligent agent to finish a race in a minimal time, and one or more additional component reward functions, providing the one or more skill components and/or the one or more personality components recites further detail on the component reward functions, without changing that the component reward functions are part of a mathematical formula of a reward function which linearly combines weights for the component reward functions, which is a mathematical concept, which is an abstract idea. Step 2A, Prong 2 - “Does the claim recite additional elements that integrate the judicial exception into a practical application?”: The limitation of training a single policy of the artificial intelligent agent over a continuous goal space including a plurality of parameterized reward functions represented by the continuous distribution of the weight for each of the plurality of component reward functions, recites mere instructions to apply training to an artificial intelligent agent policy, which does not integrate the exception into a practical application, MPEP 2106.05(d) and 2106.05(f). Step 2B - “Does the claim recite additional elements that amount to significantly more than the judicial exception?”: The limitation of training a single policy of the artificial intelligent agent over a continuous goal space including a plurality of parameterized reward functions represented by the continuous distribution of the weight for each of the plurality of component reward functions, recites mere instructions to apply training to an artificial intelligent agent policy, which is not significantly more than any recited judicial exceptions , MPEP 2106.05(f). Therefore, claim 1 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 15, Claim 15 adds the additional limitations to claim 14: further comprising improving a performance of the artificial intelligent agent over a segment of the continuous distribution of the weight by providing a skewed distribution of weight, recites mere instructions to apply performance improvement via a skewed weight distribution, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). wherein the training is performed over the skewed distribution of weight for one or more of the plurality of component reward functions recites mere instructions to apply training over a skewed weight distribution, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 15 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 16, Claim 16 adds the additional limitation to claim 15: further comprising applying the continuous distribution of weights to both a policy and a value function of a training algorithm, recites mere instructions to apply a weight distribution, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). wherein a neural network policy is updated from π(s) to π( s,ŵ ) and the action-value function is updated from Q(s, a) to Q(s, a, ŵ) by concatenating the continuous distribution of weights, ŵ with inputs related to a state, s and an action, a recites a mathematical formulas for a network policy and for an action-value function, which are mathematical concepts, which are abstract ideas. Therefore, claim 16 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 17, Claim 17 adds the additional limitations to claim 14: further comprising evaluating the single policy of the artificial intelligent agent at inference time by choosing a chosen weight for each of the plurality of component reward functions, recites an evaluation of a policy and a judgement of a weight to choose, which are mental processes, which are abstract ideas, regardless of if they’re performed on a generic computer or using generic machine learning models. wherein the artificial intelligent agent behaves optimally under a chosen reward function without any retraining recites mere instructions to apply a chosen reward function to an artificial intelligent agent , which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 17 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 18, Claim 18 recites a non-transitory computer readable storage medium which is a manufacture, with instructions for performing the function of the method of claim 1, with substantially the same limitations. Therefore the same analysis and rejection applied to claim 1 applies to claim 18. Therefore, claim 18 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 19, Claim 19 adds the additional limitation to claim 18: wherein the artificial intelligent agent is part of a racing game environment recites mere instructions to apply the artificial intelligent agent to a racing game environment, which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 19 is found to be ineligible subject matter under 35 U.S.C. 101. Regarding claim 20, Claim 20 adds the additional limitations to claim 18: further comprising evaluating the single policy of the artificial intelligent agent at inference time by choosing a chosen weight for each of the plurality of component reward functions, recites an evaluation of a policy and a judgement of a weight to choose, which are mental processes, which are abstract ideas, regardless of if they’re performed on a generic computer or using generic machine learning models. wherein the artificial intelligent agent behaves optimally under a chosen reward function without any retraining recites mere instructions to apply a chosen reward function to an artificial intelligent agent , which does not integrate the exception into a practical application, and is not significantly more than any recited judicial exceptions, MPEP 2106.05(d) and 2106.05(f). Therefore, claim 20 is found to be ineligible subject matter under 35 U.S.C. 101. Prior Art The following references are used for prior art claim rejections: Wurman et al. “ Outracing champion Gran Turismo drivers with deep reinforcement learning ” Abels et al. “ Dynamic Weights in Multi-Objective Deep Reinforcement Learning ” Cai et al. “ Random Sampling Weights Allocation Update for Deep Reinforcement Learning ” Ghosh et al. “ L earning A ctionable R epresentations with G oal -C onditioned P olicies” Pong et al. “ Skew-Fit: State-Covering Self-Supervised Reinforcement Learning ” Jaderburg et al. “ Reinforcement Learning With Unsupervised Auxiliary Tasks ” Buet-Golfouse and Pahwa “ Robust Multi-Objective Reinforcement Learning with Dynamic Preferences ” Claim Rejections - 35 USC § 103 The following is a quotation of 35 U.S.C. 103 which forms the basis for all obviousness rejections set forth in this Office action: A patent for a claimed invention may not be obtained, notwithstanding that the claimed invention is not identically disclosed as set forth in section 102, if the differences between the claimed invention and the prior art are such that the claimed invention as a whole would have been obvious before the effective filing date of the claimed invention to a person having ordinary skill in the art to which the claimed invention pertains. Patentability shall not be negated by the manner in which the invention was made. Claims 1 , 5 - 14, and 16- 20 are rejected under 35 U.S.C. 103 as being unpatentable over Wurman et al. “ Outracing champion Gran Turismo drivers with deep reinforcement learning ”, hereinafter Wurman, in view of Abels et al. “ Dynamic Weights in Multi-Objective Deep Reinforcement Learning ”, hereinafter Abels , further in view of Cai et al. “ Random Sampling Weights Allocation Update for Deep Reinforcement Learning ”, hereinafter Cai, further in view of Ghosh et al. “ L earning A ctionable R epresentations with G oal -C onditioned P olicies”, hereinafter Ghosh. Regarding claim 1, Wurman teaches A method for training an artificial intelligent agent that generalizes over continuous behaviors in multiple dimensions, ((Wurman Pg. 2) “ The core actions of the agent were mapped to two continuous-valued dimensions: changing velocity (accelerating or braking) and steering (left or right) ”, changing velocity and steering are behaviors) the method comprising: defining a reward function based on a state and an action as a linear combination of a plurality of component reward functions and a weight for each of the plurality of component reward functions; ((Wurman Pg. 7) “The reward function was a hand-tuned linear combination of reward components computed on the transition between the previous state s and current state s ′. The reward components were: course progress ( R cp ), off-course penalty ( R soc or R loc ), wall penalty ( R w ), tyre -slip penalty ( R ts ), passing bonus ( R ps ), any-collision penalty ( R c ), rear-end penalty (R r ) and unsporting-collision penalty ( R uc ). The reward weightings for the three tracks are shown in Extended Data Table 1”, Wurman Pg. 14, Extended Data Table 1 shows the weights for each reward component function) … including a plurality of parameterized reward functions … (Wurman Pg. 14, Extended Data Table 1 shows the weights for each reward component function, with each course having reward component functions parameterized with different weights) Abels teaches the following further limitations that Wurman does not explicitly teach: sampling multiple dimensions of the weight for each of the plurality of component reward [functions] (( Abels Pg. 5) “The reward vectors are N-dimensional: r = (r 1 , ..., r N ). The first N−1 elements correspond to the amount of each of the N−1 resources the agent sold, the last element is the consumed fuel…The weight vector w expresses the relative importance of the objectives, i.e., the price per resource”, weights for rewards with multiple dimensions have multiple dimensions, Wurman but not Abels teaches component reward functions) from a continuous distribution … (( Abels Pg. 6) “First, we evaluate the performance for sparse and large weight changes; the current weight, w, is randomly sampled from a Dirichlet distribution (α = 1) every 50k steps for Minecart”, Dirichlet distributions are continuous) and training a single policy of the artificial intelligent agent (( Abels Pg. 5) “ Each policy π w is trained for the active weight vector w following scalarized deep Q-learning ”) over a continuous [goal] space (( Abels Pg. 5 ) “ we propose an original benchmark, the Minecart problem. Minecart has a continuous state space, stochastic transitions and delayed rewards ”) including [a plurality of] parameterized reward functions (( Abels Pg. 2) “ Multi-Objective MDPs (MOMDP) (White & Kim, 1980) are MDPs with a vector-valued reward function r t = R( s t , a t ). Each component of r t corresponds to one objective. A scalarization function f maps the multi-objective value V π of a policy π to a scalar value, i.e., the user utility. In this paper we focus on linear f; each objective, i , is given a weight w i ”, Wurman more clearly teaches a plurality of parameterized reward functions) represented by the continuous distribution of the weight for each of the plurality of component reward [ functions ] (( Abels Pg. 6) “ First, we evaluate the performance for sparse and large weight changes; the current weight, w, is randomly sampled from a Dirichlet distribution (α = 1) every 50k steps for Minecart ”, Dirichlet distributions are continuous) At the time of filing, one of ordinary skill in the art would have motivation to combine Wurman and Abels by taking the method for training an agent for using continuous, multi-dimensional behaviors with a reward function that linearly combines weighted component reward functions, which may be parameterized , taught by Wurman , and including sampling multiple weight dimensions for each reward component from a continuous distribution , as well as training a policy over a continuous space , taught by Abels , as Abels teaches: ( Abels Pg. 1) “ most real-life problems are more naturally expressed with multiple objectives. For example, autonomous drivers need to minimize travel time and fuel consumption, while maximizing safety … we focus on the setting where the weights are linear, but not fixed. Specifically, the parameters of the scalarization function change over time. For example, if fuel costs increase, a shorter travel time could no longer be worth the increased fuel consumption ”, that is, that doing so accommodates a more flexible and widely applicable paradigm of agent use, where changes in weights must be accounted for by the agent. Such a combination would be obvious. Cai teaches the following further limitation that neither Wurman nor Abels explicitly teach es : sampling [multiple dimensions of] the weight [for each of the plurality of component reward functions] from a continuous distribution between a maximum weight and a minimum weight; ((Cai Pg. 6) “ For Random Sampling Weights Allocation, we sample the weights from the Uniform (0,1) ”, uniform distributions are continuous, Abels teaches multiple weight dimensions, Wurman teaches component reward functions) At the time of filing, one of ordinary skill in the art would have motivation to combine Wurman , Abels , and Cai by taking the method for training an agent including a policy for using continuous, multi-dimensional behaviors with a reward function that linearly combines weighted component reward functions, which may be parameterized , as well as a continuous weight distribution which is sampled in multiple dimensions , jointly taught by Wurman and Abels , and having the sample weights be between a maximum weight and a minimum weight, taught by Cai, as having weights be on a common scale with a maximum and minimum, such as between 0 and 1 or -1 and 1, is well known within the art of machine learning, providing the predictable benefit of faster model convergence, that is, the point at which a trained model has outputs with the highest accuracy. Such a combination would be obvious. Ghosh teaches the following further limitation that neither Wurman, nor Abels , nor Cai explicitly teaches: and training a single policy of the artificial intelligent agent over a continuous goal space … ((Ghosh Pg. 14) “ We train a stochastic goal-conditioned policy π(·|s, g) using TRPO with an entropy regularization term, where the goal space G coincides with the state space S. In every episode, a starting state and a goal state s, g ∈ S are sampled from a uniform distribution on states ”, (Ghosh Pg. 2) “ In RL, the goal is to learn a policy π θ ( a t |s t ) that maximizes the expected return … Typically, RL learns a single task that optimizes for a particular reward function. If we instead would like to train a policy that can accomplish a variety of tasks, we might instead train a policy that is conditioned on another input – a goal. When the different tasks directly correspond to different states, this amounts to conditioning the policy π on both the current and goal state. The policy π θ ( a t |s t , g) is trained to reach goals from the state space g ∼ S ”) At the time of filing, one of ordinary skill in the art would have motivation to combine Wurman, Abels , Cai, and Ghosh by taking the method for training an agent including a policy for using continuous, multi-dimensional behaviors with a reward function that linearly combines weighted component reward functions, which may be parameterized , as well as a continuous weight distribution which is sampled in multiple dimensions between a maximum and a minimum, jointly taught by Wurman, Abels , and Cai, and having the policy be trained over a goal space, taught by Ghosh, as doing so is well known within the art of reinforcement learning, allowing the policy to explore the goal space in order to achieve goals that maximize rewards, producing an optimal model . Such a combination would be obvious. Regarding claim 5 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 1, Abels further teaches: further comprising repeatedly re-sampling the continuous distribution of weights (( Abels Pg. 6) “First, we evaluate the performance for sparse and large weight changes; the current weight, w, is randomly sampled from a Dirichlet distribution (α = 1) every 50k steps for Minecart”, Dirichlet distributions are continuous, sampling the weights every 50k steps is repeatedly resampling) during a training rollout, (( Abels Pg. 3) “We propose our first main contribution, Conditioned Network (CN), in which a UVFA is adapted to output Q-value vectors conditioned on an input weight vector (Figure 1). The training algorithm follows the standard DQN algorithm, i.e., the agent acts ε-greedily and stores its experiences in a replay buffer from which transitions are sampled to train the network on”, training with storing experiences corresponds to training rollout) wherein the artificial intelligent agent becomes robust to reward function changes during ongoing trajectories (( Abels Pgs. 1-2) “We test the performance of our algorithms on two weight change scenarios and find that, while methods from related settings can be adapted to the dynamic weights setting, only our proposed CN can both quickly adapt to sparse abrupt weight changes and also converge to optimal policies when weight changes occur regularly. Furthermore, by maintaining a set of diverse trajectories, DER improves the performance of all tested algorithms”) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman , Abels , Cai , and Ghosh for the parent claim of claim 5 , claim 1. No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 6 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 1, Wurman further teaches: further comprising applying the [continuous] distribution of weights to both a policy and a value function of a training algorithm ((Wurman Pg. 3) “We trained GT Sophy using a new deep RL algorithm we call quantile regression soft actor-critic (QR-SAC). This approach learns a policy (actor) that selects an action on the basis of the agent’s observations and a value function (critic) that estimates the future rewards of each possible action. QR-SAC extends the soft actor-critic approach by modifying it to handle N-step returns and replacing the expected value of future rewards with a representation of the probability distributions of those rewards”, using a representation of a probability distribution of rewards for SAC, when the rewards are weighted as in Wurman, corresponds to applying a distribution of weights to the policy and value functions, Abels teaches a continuous distribution for weights) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman , Abels , Cai , and Ghosh for the parent claim of claim 6 , claim 1. No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 7 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 6 , Abels further teaches: further comprising updating a neural network policy from π(s) to π(s, ŵ) (( Abels Pg. 2) “Under the standard assumption that future rewards are discounted by a factor γ ∈ [0,1], the goal of the agent is to find a policy π*( a|s )”, ( Abels Pg. 5) “Each policy π w is trained for the active weight vector w following scalarized deep Q-learning”) and the action-value function Q(s, a) to Q(s, a, ŵ) (( Abels Pg. 2) “Correspondingly, the Q-function Q π : S × A → R maps a state-action pair to the expected return obtained from that state when the action is executed”), ( Abels Pg. 5) “Q(a, s; w) is the network’s Q-value for action a in state s and weight vector w”) by concatenating the continuous distribution of weights, ŵ with inputs related to state, s (( Abels Pg. 3) “Based on the observation that a goal is often a subset of the set of states, the network learns goal and state embeddings and uses a distance-based metric to combine both embeddings. This is achieved offline, by learning several value functions independently, factorizing embeddings and then training a network to approximate these values for any given goal. In MORL, a goal would be a specific weight vector and as such there is no clear relation between the goal (i.e., the importance of each objective) and the state. One could fall back on the concatenation of state and goal embeddings”) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman , Abels , Cai , and Ghosh for the parent claim of claim 7 , claim 6 . No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 8 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 1, Wurman further teaches: further comprising evaluating the single policy of the artificial intelligent agent at inference time ((Wurman Pg. 9) “Extended Data Figure 3 illustrates the policy-selection process. Agent policies were saved at regular intervals during training. Each saved policy then competed in a single-race scenario against other AI agents, and various metrics, such as lap times and car collisions, were gathered and used to filter the saved policies to a smaller set of candidates. These candidates were then run through an n- athlon —a set of pre-specified evaluation scenarios—testing their lap speed and performance in certain tactically important scenarios…each pair of policies competed in a multi-race, round-robin, policy-versus-policy tournament. These competitions were scored using the same team scoring as that in the exhibition event and evaluated on collision metrics. From these results, the committee chose policies that seemed to have the best chance of winning against the human drivers while minimizing penalties”, evaluation scenarios correspond to evaluations at inference time) by choosing a chosen weight for each of the plurality of component reward functions, ((Wurman Pg. 7) “The reward function was a hand-tuned linear combination of reward components computed on the transition between the previous state s and current state s ′. The reward components were: course progress ( R cp ), off-course penalty ( R soc or R loc ), wall penalty ( R w ), tyre -slip penalty ( R ts ), passing bonus ( R ps ), any-collision penalty ( R c ), rear-end penalty (R r ) and unsporting-collision penalty ( R uc ). The reward weightings for the three tracks are shown in Extended Data Table 1”, Wurman Pg. 14, Extended Data Table 1 shows the weights for each reward component function, minimized penalties are chosen weights, as most of the weighted component reward functions are penalties) wherein the artificial intelligent agent behaves accordingly under a chosen reward function ((Wurman Pg. 7) “The reward function was a hand-tuned linear combination of reward components computed on the transition between the previous state s and current state s ′”) without any retraining ((Wurman Pg. 7) “To train policies that could drive in traffic, 21 rollout workers were used for between 7 and 12 days. In both cases, one worker was primarily evaluating intermediate policies, rather than generating new training data”, evaluating intermediate policies without generating new training data corresponds to behaving under the policy without retraining) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman , Abels , Cai , and Ghosh for the parent claim of claim 8 , claim 1 . No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 9 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 1, Wurman further teaches: wherein the artificial intelligent agent operates in a racing game environment ((Wurman Abstract) “Racing simulations, such as the PlayStation game Gran Turismo, faithfully reproduce the non-linear control challenges of real race cars while also encapsulating the complex multi-agent interactions. Here we describe how we trained agents for Gran Turismo that can compete with the world’s best e-sports drivers”) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman , Abels , Cai , and Ghosh for the parent claim of claim 9 , claim 1 . No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 10 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 9, further comprising: Wurman further teaches: providing a base reward as one of the plurality of component reward functions, ((Wurman Pg. 7) “the primary reward component rewarded the amount of progress made along the track since the last observation”) the base reward motivating the artificial intelligent agent to finish a race in a minimal time; ((Wurman Pg. 3) “The agent was given a progress reward for the speed with which it advanced around the track and penalties if it went out of bounds, hit a wall or lost traction. These shaping rewards allowed the agent to quickly receive positive feedback for staying on the track and driving fast”, positive feedback for driving fast corresponds to motivation to finish a race in minimal time) and providing one or more additional ones of the plurality of component reward functions to provide one or more skill component reward functions and/or one or more personality component reward functions ((Wurman Pg. 8) “ Tyre slip makes it more difficult to control the car. To assist learning, we included a penalty when the tyres were slipping in a different direction from where they were pointing”, a penalty (i.e. a negative award) for tire slipping which reduces ability to control a car corresponds to a skill component reward function) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman, Abels , Cai, and Ghosh for the parent claim of claim 10 , claim 9 . No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 11 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 10, Wurman further teaches: wherein the [continuous distribution of] weights for the one or more additional ones describe an importance of each of the one or more additional ones in relation to a fixed weight for the base reward (Wurman Pg. 14, Extended Data Table 1 shows the weights for each reward component function, which vary based on importance, however all the weights are fixed, Abels teaches a continuous distribution of weights) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman, Abels , Cai, and Ghosh for the parent claim of claim 11 , claim 10 . No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 12 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 10, Wurman further teaches: wherein the one or more additional ones of the plurality of component reward functions include at least one of the following: (a) a penalty for degradation of car tires of the artificial intelligent agent during an environment step; (b) a penalty for a fuel use by the artificial intelligent agent during an environment step; (c) a linear penalty for tire slip ratios and angles; ((Wurman Pg. 8) “ Tyre slip makes it more difficult to control the car. To assist learning, we included a penalty when the tyres were slipping in a different direction from where they were pointing: [Equation], where s tsr,i is the tyre -slip ratio for the ith tyre and s tsθ,i is the angle of the slip from the forward direction of the ith tyre ”) (d) a linear positive reward for tire slip ratio and angle; (e) an edge distance penalty that linearly increases with a proximity of the artificial intelligent agent to an edge of a racing track; (f) a set of reward parts penalizing the artificial intelligent agent for driving in corresponding slices of the racing track defined by a distance to a centerline thereof; (g) a passing reward with independently weighted positive and negative parts for overtaking and being overtaken by other cars, respectively; (h) a penalty on a change in steering angle during an environment step; ( i ) a penalty for colliding with other vehicles; and (j) a penalty for a car of the artificial intelligent agent driving off course At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman, Abels , Cai, and Ghosh for the parent claim of claim 12 , claim 10 . No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 13 , Wurman, Abels , Cai , and Ghosh jointly teach The method of claim 1 2 , Wurman further teaches: wherein each of the one or more additional ones of the plurality of component reward functions are defined within the single policy of the artificial intelligent agent ((Wurman Pg. 3) “ We trained GT Sophy using a new deep RL algorithm we call quantile regression soft actor-critic (QR-SAC). This approach learns a policy (actor) that selects an action on the basis of the agent’s observations and a value function (critic) that estimates the future rewards of each possible action ”, (Wurman Pg. 7) “ The reward function was a hand-tuned linear combination of reward components computed on the transition between the previous state s and current state s ′ ”, a policy that selects actions based on future rewards, provided via combination of component rewards, corresponds to the component reward functions being defined within the policy) At the time of filing, one of ordinary skill in the art would have motivation to combine the method jointly taught by Wurman, Abels , Cai, and Ghosh for the parent claim of claim 13 , claim 12 . No new embodiments are introduced, so the reason to combine is the same as for the parent claim. Regarding claim 1 4 , Wurman teaches A method for providing an artificial intelligent agent in a racing game that is tunable to one or more skill components and/or one or more personality components, ((Wurman Pg. 8) “ Tyre slip makes it more difficult to control the car. To assist learning, we included a penalty when the tyres were slipping in a different direction from where they were pointing”, a penalty (i.e. a negative award) for tire slipping which reduces ability to control a car corresponds to a skill component reward function) the method comprising: defining a reward function based on a state and an action as a linear combination of a plurality of component reward functions and a weight for each of the plurality of component reward functions; ((Wurman Pg. 7) “The reward function was a hand-tuned linear combination of reward components computed on the transition between the previous state s and current state s ′. The reward components were: course progress ( R cp ), off-course penalty ( R soc or R loc ), wall penalty ( R w ), tyre -slip penalty ( R ts ), passing bonus ( R ps ), any-collision penalty ( R c ), rear-end penalty (R r ) and unsporting-collision penalty ( R uc ). The reward weightings for the three tracks are shown in Extended Data Table 1”, Wurman Pg. 14, Extended Data Table 1 shows the weights for each reward component function) … including a plurality of parameterized reward functions … (Wurman Pg. 14, Extended Data Table 1 shows the weights for each reward component function, with each course having reward component functions parameterized with different weights) w herein the plurality of component reward functions include a base reward, ((Wurman Pg. 7) “the primary reward component rewarded the amount of progress made along the track since the last observation”) motivating the artificial intelligent agent to finish a race in a minimal time , … ((Wurman Pg. 3) “The agent was given a progress reward for the speed with which it advanced around the track and penalties if it went out of bounds, hit a wall or lost traction. These shaping rewards allowed the agent to quickly receive positive feedback for staying on the track and driving fast”, positive feedback for driving fast corresponds to motivation to finish a race in minimal time) … and one or more additional component reward functions, providing the one or more skill components and/or the one or more personality components ((Wurman Pg. 8) “ Tyre slip makes it more difficult to control the car. To assist learning, we included a penalty when the tyres were slipping in a different direction from where they were pointing”, a penalty (i.e. a negative award) for tire slipping which reduces ability to control a car corresponds to a skill component reward function) Abels teaches the following further limitations that Wurman does not explicitly teach: sampling multiple dimensions of the weight for each of the plurality of component reward [functions] (( Abels Pg. 5) “The reward vectors are N-dimensional: r = (r 1 , ..., r N ). The first N−1 elements correspond to the amount of each of the N−1 resources the agent sold, the last element is the consumed fuel…The weight vector w expresses the relative importance of the objectives, i.e., the price per resource”, weights for rewards with multiple dimensions have multiple dimensions, Wurman but not Abels teaches component reward functions) from a continuous distribution … (( Abels Pg. 6) “First, w