System and method for safety and efficacy override of an autonomous system
Described is a system for competency assessment of an autonomous system. The system extracts semantic concepts representing a situation. Actions taken by the autonomous system are associated with semantic concepts that are activated when the actions are taken in the situation. The system measures an outcome of the actions taken in the situation and generates a reward metric. The semantic concepts representing the situation are stored as a memory with the actions taken in the situation and the reward metric as a memory. A prospective simulation is generated based on recall of the memory. A competency metric and an experience metric are determined. Competent operational control of the autonomous system is maintained when at least one of the competency metric and the experience metric is above a minimum value. An alert is generated when at least one of the competency metric and the experience metric falls below the minimum value.
1. A system for safety and efficacy override of an autonomous system, the system comprising:
one or more processors and a non-transitory computer-readable medium having executable instructions encoded thereon such that when executed, the one or more processors perform operations of:
extracting a plurality of semantic concepts representing a situation;
associating actions taken by the autonomous system with a set of semantic concepts that are activated when the actions are taken in the situation;
using task-specific reward functions of the autonomous system's performance, measuring an outcome of the actions taken in the situation and generating a reward metric;
storing the plurality of semantic concepts representing the situation together with the actions taken in the situation and the reward metric as sequential episodic experiences;
decomposing a new situation into a set of semantic concepts;
performing a series of stochastic prospective simulations using the set of semantic concepts and the stored sequential episodic experiences;
determining a measure of competence of the autonomous system for handling the new situation based on the series of stochastic prospective simulations;
maintaining competent operational control of the autonomous system when the measure of competence is above an established minimum allowable value; and
generating an alert when the measure of competence falls below the established minimum allowable value.
2. The system as set forth in claim 1 , wherein the one or more processors further perform an operation of determining a depth-of experience estimation (DoEE), wherein the DoEE is an estimate of how often an action has been taken by the autonomous system in a given situation.
3. The system as set forth in claim 1 , wherein the measure of competence is a Receiver-Operator Characteristic (ROC) curve-based measure of competence.
4. The system as set forth in claim 1 , wherein the one or more processors further perform an operation of initiating handover of control of the autonomous system to a human user when the measure of competence falls below the established minimum allowable value.
5. The system as set forth in claim 1 , wherein the one or more processors further perform an operation of initiating handover of control from a human user to an assistive system when the measure of competence falls below the established minimum allowable value.
6. The system as set forth in claim 1 , wherein the autonomous system is an autonomous driving system.
7. A computer implemented method for safety and efficacy override of an autonomous system, the method comprising an act of:
causing one or more processers to execute instructions encoded on a non-transitory computer-readable medium, such that upon execution, the one or more processors perform operations of:
extracting a plurality of semantic concepts representing a situation;
associating actions taken by the autonomous system with a set of semantic concepts that are activated when the actions are taken in the situation;
using task-specific reward functions of the autonomous system's performance, measuring an outcome of the actions taken in the situation and generating a reward metric;
storing the plurality of semantic concepts representing the situation together with the actions taken in the situation and the reward metric as sequential episodic experiences;
decomposing a new situation into a set of semantic concepts;
performing a series of stochastic prospective simulations using the set of semantic concepts and the stored sequential episodic experiences;
determining a measure of competence of the autonomous system for handling the new situation based on the series of stochastic prospective simulations;
maintaining competent operational control of the autonomous system when the measure of competence is above an established minimum allowable value; and
generating an alert when the measure of competence falls below the established minimum allowable value.
8. The method as set forth in claim 7 , wherein the one or more processors further perform an operation of determining a depth-of experience estimation (DoEE), wherein the DoEE is an estimate of how often an action has been taken by the autonomous system in a given situation.
9. The method as set forth in claim 7 , wherein the measure of competence is a Receiver-Operator Characteristic (ROC) curve-based measure of competence.
10. The method as set forth in claim 7 , wherein the one or more processors further perform an operation of initiating handover of control of the autonomous system to a human user when the measure of competence falls below the established minimum allowable value.
11. The method as set forth in claim 7 , wherein the one or more processors further perform an operation of initiating handover of control from a human user to an assistive system when the measure of competence falls below the established minimum allowable value.
12. The method as set forth in claim 7 , wherein the autonomous system is an autonomous driving system.
13. A computer program product for safety and efficacy override of an autonomous system, the computer program product comprising:
computer-readable instructions stored on a non-transitory computer-readable medium that are executable by a computer having one or more processors for causing the processor to perform operations of:
extracting a plurality of semantic concepts representing a situation;
associating actions taken by the autonomous system with a set of semantic concepts that are activated when the actions are taken in the situation;
using task-specific reward functions of the autonomous system's performance, measuring an outcome of the actions taken in the situation and generating a reward metric;
storing the plurality of semantic concepts representing the situation together with the actions taken in the situation and the reward metric as sequential episodic experiences;
decomposing a new situation into a set of semantic concepts;
performing a series of stochastic prospective simulations using the set of semantic concepts and the stored sequential episodic experiences;
determining a measure of competence of the autonomous system for handling the new situation based on the series of stochastic prospective simulations;
maintaining competent operational control of the autonomous system when the measure of competence is above an established minimum allowable value; and
generating an alert when the measure of competence falls below the established minimum allowable value.
14. The computer program product as set forth in claim 13 , further comprising instructions for causing the one or more processors to further perform an operation of determining a depth-of experience estimation (DoEE), wherein the DoEE is an estimate of how often an action has been taken by the autonomous system in a given situation.
15. The computer program product as set forth in claim 13 , wherein the measure of competence is a Receiver-Operator Characteristic (ROC) curve-based measure of competence.
16. The computer program product as set forth in claim 13 , further comprising instructions for causing the one or more processors to further perform an operation of initiating handover of control of the autonomous system to a human user when the measure of competence falls below the established minimum allowable value.
17. The computer program product as set forth in claim 13 , further comprising instructions for causing the one or more processors to further perform an operation of initiating handover of control from a human user to an assistive system when the measure of competence falls below the established minimum allowable value.
18. The computer program product as set forth in claim 13 , wherein the autonomous system is an autonomous driving system.
19. The system as set forth in claim 1 , wherein the series of stochastic prospective simulation are performed for each of a task objective and a lure objective, thereby yielding target and lure distributions of performance estimates for a given action.
20. The system as set forth in claim 19 , wherein the one or more processors further perform operations of:
generating a Receiver-Operator Characteristic (ROC) curve from each of the target and lure distributions, wherein the target distribution represents true positive values and the lure distribution represents false positive values; and
determining an Area Under the Curve (AUC) metric for the target and lure distributions, wherein the AUC metric represents competency of the autonomous system in the new situation.