Responses to vulnerable road user's adversarial behavior

A system for responding to adversarial behavior within an autonomous vehicle, comprising a pedestrian detection system in communication with a vehicle controller and an adversarial intent algorithm adapted to:

• determine that a risk level of adversarial behavior from at least one pedestrian within proximity of the autonomous vehicle is one of no risk of adversarial behavior, low risk of adversarial behavior or high risk of adversarial behavior by:
• analyzing, with a crossing behavior algorithm, crossing behaviors of the at least one pedestrian and determining if crossing behaviors of the at least one pedestrian are adversarial based on whether the autonomous vehicle has the right of way, the type of crossing including any signs, lights or signals present, and how long the at least one pedestrian takes to complete the crossing;
• analyzing, with a non-verbal cue algorithm, non-verbal cues displayed by the at least one pedestrian and determining if non-verbal cues displayed by the at least one pedestrian are adversarial based on comparing captured images and body gestures and actions of the at least one pedestrian to known patterns of adversarial behavior;
• analyzing, with an audible cues algorithm, audible cues displayed by the at least one pedestrian and determining if audible cues displayed by the at least one pedestrian are adversarial based on comparing captured acoustic signals from the at least one pedestrian to known sounds that indicate adversarial behavior;
• and
• combining, with an information fusion algorithm, information from the crossing behavior algorithm, the non-verbal cue algorithm and the audible cues algorithm;
• and
• provide actions to be performed by the autonomous vehicle in response to the adversarial behavior based on the risk level of the adversarial behavior.

A system for responding to adversarial behavior directed toward an autonomous vehicle, comprising a pedestrian detection system in communication with a vehicle controller and an adversarial intent algorithm adapted to:

• determine that a risk level of adversarial behavior from at least one pedestrian within proximity of the autonomous vehicle is one of no risk of adversarial behavior, low risk of adversarial behavior or high risk of adversarial behavior by:
• analyzing, with a crossing behavior algorithm, crossing behaviors of the at least one pedestrian and determining if crossing behaviors of the at least one pedestrian are adversarial based on whether the autonomous vehicle has the right of way, the type of crossing including any signs, lights or signals present, and how long the at least one pedestrian takes to complete the crossing:
• analyzing, with a non-verbal cue algorithm, non-verbal cues displayed by the at least one pedestrian and determining if non-verbal cues displayed by the at least one pedestrian are adversarial based on comparing captured images and body gestures and actions of the at least one pedestrian to known patterns of adversarial behavior;
• analyzing, with an audible cues algorithm, audible cues displayed by the at least one pedestrian and determining if audible cues displayed by the at least one pedestrian are adversarial based on comparing captured acoustic signals from the at least one pedestrian to known sounds that indicate adversarial behavior;
• and
• combining, with an information fusion algorithm, information from the crossing behavior algorithm, the non-verbal cue algorithm and the audible cues algorithm;
• determine if the adversarial behavior is persistent;
• and
• provide actions to be performed by the autonomous vehicle in response to the adversarial behavior based on the risk level of the adversarial behavior;
• wherein:
• when the adversarial intent algorithm determines that there is no risk of adversarial behavior, the vehicle controller is adapted to provide no actions to be performed responsive to adversarial behavior by the autonomous vehicle;
• when the adversarial intent algorithm determines that there is low risk of adversarial behavior, the vehicle controller is adapted to initiate at least one of visual and auditory warnings to the pedestrian;
• when the adversarial behavior continues for less than a pre-determined time window, the vehicle controller is adapted to stop the at least one of visual and auditory warnings, and reset the classification, by the adversarial intent algorithm, of the adversarial behavior risk level;
• when the adversarial behavior continues for more than the pre-determined time window, the vehicle controller is adapted to cause the autonomous vehicle to at least one of back away from the adversarial behavior or re-route to avoid the adversarial behavior;
• when the adversarial intent algorithm determines that there is high risk of adversarial behavior, the vehicle controller is adapted to report the high risk of adversarial behavior to at least one of vehicle assistance service providers, authorities, infrastructure services providers or crowd-sensing services;
• when the adversarial behavior continues for less than a third pre-determined time window, the vehicle controller is adapted to stop the reporting to at least one of vehicle assistance service providers, authorities, infrastructure services providers or crowd-sensing services, and, when the adversarial behavior continues for more than a pre-determined time window, the vehicle controller is adapted to at least one of:
• cause the autonomous vehicle to at least one of:
• back away from the adversarial behavior, or, re-route to avoid the adversarial behavior;
• cause control of the autonomous vehicle to be transferred from the vehicle controller to an eligible occupant within the autonomous vehicle;
• or
• cause control of the autonomous vehicle to be transferred from the vehicle controller to a remotely located tele-driving entity.