Lecture, Interview
Sterling Anderson, Co-Founder, Aurora - MIT Self-Driving Cars
- Sterling Anderson, former MIT PhD researcher and former head of Tesla Autopilot, co-founded Aurora in December 2016 to develop a new self-driving architecture leveraging deep learning, high-resolution radar, and LiDAR.
- Aurora has secured partnerships with Volkswagen Group and Hyundai Motor Company to develop and scale autonomous platforms across their global vehicle fleets.
- Aurora maintains physical operations and test fleets in Palo Alto, California, and Pittsburgh, Pennsylvania.
- The company has adopted a non-automotive manufacturing business model, focusing on developing the software stack while relying on partners for vehicle production and distribution.
- Aurora currently plans to run pilot programs with Waymo shuttles, distinguishing these pilots from broad commercialization.
- The "intelligent co-pilot" project (developed circa 2007 at MIT) utilized constraint-based control and homotopy planning to keep vehicles within safe state-space tubes rather than fixed paths.
- In unmanned ground vehicle testing, the co-pilot system reduced collision incidents by 72% compared to manual control while simultaneously increasing operational speed by approximately 20–30%.
- Test subjects reported a statistically significant 12% increase in perceived control when the co-pilot was engaged, despite the system assuming 43% of the actual control authority.
- Anderson's research indicated that torque feedback mechanisms in steering wheels must be calibrated based on driver experience to avoid confusing mental models.
- Aurora's core team includes Chris Hermson (ex-Google self-driving), Drew Bagnall (ex-Uber autonomous perception and CMU professor), and Anderson.
- The company believes the current technological bottleneck lies in forecasting human intent and future behaviors (perception and prediction) rather than raw sensor capability.
- Aurora's economic strategy prioritizes deploying high-performance, multi-modal sensor systems (including LiDAR) for initial fleet deployment, with costs expected to decline as volumes increase and supply chains mature.
- The company views security against malicious actors as an ongoing "cat and mouse" dynamic requiring constant architectural updates rather than a one-time solution.
- Anderson anticipates self-driving technology will enable new vehicle interiors, reduced passive safety features (e.g., crumple zones), and new revenue streams via in-vehicle services like AR and location data.
- Aurora acknowledges that widespread adoption will cause job displacement in the transportation sector, advocating for proactive planning to transition affected workers into new roles.
- Anderson predicts that while Vehicle-to-Everything (V2X) communication is beneficial, it will not significantly impact short-term deployment due to slow infrastructure rollout and scaling challenges.
- The company views the convergence of electrification, ride-sharing, and autonomy as the primary vector for maximizing societal efficiency and roadway utilization.
- Anderson plans to deploy technology only after statistical evaluations establish safety levels that exceed the average human driver, acknowledging that exhaustive scenario testing is impossible.