Escaping the Autonomous Vehicle Envirofence
Geoff Deane, CTO
More than eighty cities across the globe are piloting AVs. A majority of these pilots deploy self-driving systems in near perfect weather conditions to train the AV.
This is concerning as it means the multi-trillion-dollar industry solution to everyday environmental conditions is to avoid them altogether. We’re envirofencing the AV.
It’s no secret that Mother Nature has been a challenge for AV engineering teams. AV engineers have been facing a mountain of technical hurdles, but tackling adverse environmental conditions requires specific engineering focus. Vehicles driven today enjoy a century of development across dozens of technologies that ensure we can safely see when we drive.
Addressing AV blockages (e.g., snow, mud, and insects) of tomorrow’s perception tech (e.g., cameras, LiDAR, and other hardware sensors) must be strategic and approach the challenge from all angles: avoidance, mitigation, and removal. Our industry doesn’t know which tactic will get the job done. Each prototype AV system is different and proprietary, so we’re slowly collecting data in silos. Until we have decades worth of driverless data to analyze — in real environmental conditions — safety necessitates doing everything we can.
1. Blockage Avoidance: These technologies prevent things like freezing rain, mud, bugs, and other environmental conditions from occluding or blocking perception surfaces. Future AVs will shield the hardware sensors from the environment when possible and will use advanced coatings to reduce blockages when not.
2. Blockage Impact Mitigation: When a perception surface gets blocked, how long can the AV can continue driving safely? AVs use multiple cameras and redundant hardware sensors to stitch together the real-time 3-D driving environment. Longer-term, as Petty notes, connectivity will be key. We’ll see AVs communicating with each other, buses, trains, scooters, stoplights, pedestrian crossing lanes, bikes, etc., to get an even better understanding of their current environment. In general, processing more information with greater redundancy will lengthen the amount of time an AV can operate safely with compromised vision.
3. Blockage Removal: Eventually a blocked surface will need to be cleaned off. Human hands can’t intervene every time a sensor gets dirty. For one, humans may not always be on-board. To keep vehicles on the road, we need technologies that communicate with the AV system, understand when a perception surface is blocked, and clean the blocked surface.
Every AV team is struggling to understand the severity that perception blockage issues pose to the safety of an AV. Safety mandates we think strategically, and comprehensively. Seeing the way through the AV envirofence will take focus.
