Object Detection for Obstacle Avoidance
For an autonomous vehicle to operate safely in any environment, it needs to be able to detect and locate other objects in its vicinity—the farther away, the better. Beyond the need for localization to stay on course, the car needs to “see” to perceive obstacles and avoid them. At 80 mph on a slightly wet road, a car needs to detect obstacles at least 300 meters away to have enough reaction time to stop, but commercially available automotive sensors “see” fewer than 100 meters. Draper is developing sensor technologies to detect objects at greater distances—including ones partially or fully obscured—and even predict their trajectories so that a self-driving car has more time to stop or maneuver.
To improve the performance of light detection and ranging (LiDAR) technology, Draper developed LiDAR-on-a-Chip, which provides higher resolution and further range thanks to an innovative MEMS solid-state beamsteering capability. Not only is LiDAR-on-a-chip a much smaller form factor than the current industry standard for autonomous vehicles, but the cost to produce it is an order of magnitude smaller.
To enable LiDAR technology to see through obscurants, such as fog, rain and snow, Draper designed a detector called Hemera. This all-weather detection hardware and software integrates easily with most LiDAR architectures. Hemera has been proven at up to 200 meters in rainy weather.
In addition to improving sensor technology for direct field of view, Draper is addressing the risk posed by pedestrians, bicyclists or other cars out of view emerging from behind an object suddenly into a car’s path. Draper’s PathScout vehicle sensor can use anonymized cell phone GPS data to detect unseen people and predict their immediate trajectories. This capability can work alongside Draper’s LiDAR-on-a-Chip, Hemera and other innovations to improve object detection and obstacle avoidance for autonomous vehicles.
Related News
Related Press
Access Our Capabilities
We push boundaries and deliver the capabilities you need.