Unlike automobiles, personal mobilities are designed to operate in pedestrian spaces such as sidewalks. Pedestrians and wheelchair users are expected to coexist in pedestrian spaces, and bicycle users are also expected to coexist in pedestrian spaces, albeit under certain conditions. Therefore, it is required to move without colliding with the surrounding traffic participants. In addition, unlike driving a car on a roadway, there are no strict traffic rules for moving in a walking space. In related fields such as autonomous mobile robots, control algorithms have been proposed that sense the position and speed of surrounding traffic participants and always slow down and stop to avoid collisions, and control algorithms that move without slowing down to weave through the small spaces between surrounding traffic participants. However, unlike autonomous robots, personal mobility vehicles have passengers, and the passengers and surrounding traffic participants may not accept locomotion style that does not harmonize with the surrounding traffic participants.
For example, the former algorithm, which always decelerates and stops the vehicle, makes the driver feel as if he is always being "reserved" to his surroundings. In addition, this kind of movement tends to stagnate the traffic flow, which may give the passenger the impression that he makes troubles. Therefore, such a control algorithm may undermine the desire to go out. On the other hand, the latter control algorithm, in which the vehicle moves through the surrounding traffic without decelerating, may be perceived as a dangerous traffic participant by the surrounding traffic participants, and may not be accepted by them. Therefore, it is necessary to develop a vehicle control algorithm for personal mobility that can move in harmony with other traffic participants in a walking space.
In this study, we developed a control algorithm that estimates the travel modes of traffic participants moving in front of the vehicle using cameras and laser sensors mounted on the personal mobility, and changes the collision avoidance strategy according to the estimated results. Specifically, the system determines whether the oncoming traffic participant is a pedestrian, a bicyclist, or a wheelchair user by processing the information acquired by the onboard camera. In the case of pedestrians, the system estimates the locomotion speed based on the time-series information acquired by the laser sensor, and estimates whether the pedestrian is moving quickly like a young person or slowly like an elderly person. Then, based on this information, the traffic mode of the oncoming traffic participant is determined, and collision avoidance is achieved in a way that conforms to implicit social conventions. In this way, by realizing autonomous locomotion in harmony with the surroundings, it is expected that passengers can feel social participation on an equal footing with their surroundings, thereby maintaining their mental health.
Case 1: When the oncoming traffic participant is walking slowly (assumed to be an elderly person), the own vehicle is considered to have relatively higher locomotion performance than the opponent , so the yaw motion of the own vehicle is controlled to perform active avoidance.
Case 2: When the oncoming traffic participant is walking quickly (assumed to be a young person), the vehicle does not exert itself, but instead performs passive avoidance by slowing down slightly and encouraging the opponent to avoid the vehicle.
Case 3: When the oncoming traffic participant is a cyclist, the vehicle does not exert itself, but instead slows down slightly to encourage the other party to avoid the oncoming traffic, thereby performing passive avoidance.
Case 4: When the oncoming traffic participant is a wheelchair user, the own vehicle is considered to have relatively higher locomotion performance than the opponent , so the yaw motion of the own vehicle is controlled to perform active oncoming avoidance.
Ito, Takuma, and Minoru Kamata. “Interactive Collision Avoidance Based on Surrounding Mobility Type for an Intelligent Powered Wheelchair.” International Journal of Advanced Robotic Systems, May 2013, doi:10.5772/56450.