Guided Direct Time-of-Flight Lidar Using Stereo Cameras for Enhanced Laser Power Efficiency

Area of Science:

• Robotics and Autonomous Systems
• Computer Vision and Sensor Fusion
• Photonics and Optical Engineering

Background:

• Self-driving vehicles require advanced depth-sensing technologies like Lidar.
• Conventional mechanical Lidar faces limitations in reliability, cost, and speed.
• Solid-state Lidar using direct time-of-flight (dToF) struggles with data volume for long-distance sensing.

Purpose of the Study:

• To introduce a novel 'guided' dToF approach for efficient solid-state Lidar.
• To enable power and data-efficient long-distance depth sensing for autonomous vehicles.
• To reduce the computational and energy demands of Lidar systems.

Main Methods:

• Developed a guided dToF system using a 64x32 macropixel dToF sensor.
• Integrated vision cameras to provide external depth estimates for guidance.
• Dynamically adjusted pixel exposure time windows based on external sensor data.

Main Results:

• Demonstrated outdoor scene capture at 3 frames per second up to 75 meters.
• Achieved over twenty-fold reduction in on-chip data compared to full histogram dToF.
• Reduced laser cycles by at least six times versus partial histogram approaches.
• Showcased mitigation of multipath reflections.

Conclusions:

• Guided dToF offers a power and data-efficient solution for solid-state Lidar.
• This technology enhances depth sensing capabilities for self-driving vehicles.
• Leveraging existing sensor data unlocks new possibilities for Lidar systems.