Evan Honnold

I work on motion-planning algorithms for self-driving cars. My interests include sampling-based planning, graph traversal and discrete search, parallelized algorithms, convex optimization, constraint programming, and hefty GPUs. I also enjoy long-form journalism, middle-distance running, and short-term procrastination. GitHub; LinkedIn.

Work and Education

2021
Motion Planning, Aurora Innovation (autonomous trucks)
2019
Motion Planning, Uber Advanced Technologies Group (autonomous cars) In late 2020, Uber sold the division to Aurora.
2018
MS, Robotics, Dartmouth College (Advisor: Devin Balkcom)
2014
BA, Dartmouth College
Double Major: Computer Science, Government
In college, before I discovered the joys of robotics, I interned at a law firm, a political research committee, a private-equity partnership, and a high-speed trading firm.
Sidwell Friends School

Theta-Star Planner with Angle Propagation

This discrete-search algorithm, published by Daniel, Nash, Koenig and Felner in 2010, is an extension of A*. By maintaining angle ranges (shown in green) of unobstructed space, it can avoid some collision checks. Implemented in Python, over a number of spare weekends during my time at Uber ATG; here is the code.

Planner Competition: Expensive Edge Evaluation

In some graph-based motion-planning problems, the performance bottleneck is edge evaluations -- for instance, if edges represent configuration-space motions of a robot arm and we must use forward kinematics to check them for collisions. This competition, inspired by Dellin and Srinivasa's 2016 LazySP paper, provides a way to test and compare algorithms in an environment with strictly limited edge evaluations.

Voyager 2 Trajectory Animation

This webpage (click for animation) shows the path of the Voyager 2 space probe as it uses gravity assists to visit Jupiter, Saturn, Uranus, and Neptune. I made it during the vacation between two terms of college in order to learn web development.