May Mobility has launched its third-generation autonomous driving system. With it, the AV technology company hopes to accelerate its goal of conducting driverless commercial operations that improve safety, efficiency, and rider satisfaction.

The system’s new generation offers improvements in areas such as speed, tele-assist capabilities, and detection accuracy. May Mobility expects that these improvements will benefit communities and riders alike in urban, suburban, and even rural areas. Its autonomy stack includes a third-generation vision stack and next-generation pedestrian modeling that, among other improvements, allow the vehicle to navigate efficiently across many scenarios and interactions.

The newly improved pedestrian modeling feature enables May Mobility’s autonomous system to make more robust predictions about pedestrian behaviors in complex environments. This leads the vehicle to operate more efficiently in crowded environments and respond to pedestrians who are jaywalking or violating the rules of the road. In such environments, new movement policies added to the company’s Multi-Policy Decision Making (MPDM) technology allow the vehicle to adjust its trajectory within the lane to avoid minor impediments.

Also rolled out in the third-generation system are enhanced tele-assist capabilities, which help the vehicle navigate unexpected driving encounters (such as pop-up construction sites). May Mobility’s approach to navigating such encounters pairs its MPDM technology with a remote human operator. While the vehicle moves autonomously throughout this process, the human operator monitors its surrounding environment – ensuring it remains within its operating parameters while approving actions proposed by the vehicle itself. The tele-assist operator can also provide hints to the autonomy system that are used to generate alternative routes through a complex situation.

An upgraded version of the company’s on-vehicle traffic light detection offers increased precision in challenging lighting conditions, such as early mornings and late evenings. The feature also detects non-working traffic lights, leading the vehicle to treat them as stop points.