We’ve taken advantage of some bad weather recently at our test track to check out our lidar filtering software for object detection in bad weather. Our newer test car, “Rover”, ran through tests to detect vehicles and pedestrians in heavy raining conditions. While not crossing every hurdle we’d like at higher speeds, we detected and avoided all of those obstacles at local road speeds of up to 30 MPH.
Autonomy for all weather conditions
This kind of capability is not easy – which is why so many AV companies are located in California or Nevada. It is convenient to run only in good weather but every now and then you need to try out the broader weather spectrum and see how you do. We did very well and we have yet to turn on our specialized filtering which is also running 24×7 at a commercial deployment on the Pennsylvania Turnpike. That should improve our performance to where we need it to be.
Besides filtering, other factors do come into play – the effectiveness of any Lidar sensor at longer distances isn’t helped by bad weather. So a sensor that performs well but that doesn’t have a lot of extra range will also have more issues in nasty weather. Further, when considering snow and sleet, there will need to be a way to keep sensor windows clear whether from dedicated wipers to heated screens, etc. There’s a reason so many automakers put their lane-detection cameras high up behind the windscreen and inside the cockpit!
We are also working on closing some new deals requiring advanced path management features that will come in handy in slippery conditions. These features require us to monitor our projected and actual path and report when those paths are starting to deviate. In this example, the concern is slipping in mud or gravel, but this work can also apply to cornering at speed on normal roads. Our test car Rover is well suited to try these conditions and soon enough we’ll be trying those too. In the meantime, rain testing does make for a clean car!