Hao-Tien (Lewis) Chiang

Here's a short bio and introduction to my research:

I got my BS degree in Atmospheric Sciences at the National Taiwan University. I first joined the University of New Mexico as a PhD student in Physics and worked on developing Quantum Algorithms for 3 years. Due to the love for robotics and coding (I learned Java by myself and object oriented programming blew my mind), I then transferred to the PhD program in Computer Science at UNM in Spring 2015.

Since joining CS, I've been working on bring robots to our everyday life. To do so, we need novel algorithms to generate safe and efficient robot motions (so it doesn't flip tables or run into people) in dynamic, noisy and unstructured environments. I was very lucky since I learned a bunch in motion planning and control theory at Tapia Lab and machine learning at Google Brain Robotics. This allows me to combine tools and concepts from these rich, yet drastically different paradigms.

In one example, we combined stochastic reachability from control theory with artificial potential fields and sampling-based planning from motion planning. This allowed us to systematically reason about how to generate robot motion in the presence of obstacle motion uncertainty [Chiang et. al. T-RO 17, Chiang et. al. ICRA 17].

In another example, we combined machine learning with motion planning by approximating the expensive swept volume computations by deep neural networks. This revolutionary distance measure significantly increases the efficiency and solution quality of sampling-based motion planning [Chiang et. al. WAFR 18].

Lastly, thanks to my experience at Google Brain, we combined all three paradigms in RL-RRT [Chiang et. al. RA-L 18]. This method uses deep reinforcement learning to navigate locally for robots with complex dynamics (e.g., spacecrafts and surface ships) in noisy environments. Next, we used sampling-based motion planners to guide the learned policy in order to achive rapid, global exploration. This allowed us to automatically and efficiently generate safe motions for robots with complex dynamics.

Overall, by combining the three paradigms, we were able to address many important challenges that currently limit robots to controlled lab spaces. We aim to continue combining Learning, Planning and Control in order to address major robotics issues such as the sim-to-real gap, data-efficiency, uncertainty, social compliance and trust worthiness.

After joining Waymo Researchin 2020, I collaborated extensively with Behavior Prediction (of other cars and road participants) and Motion Planning teams. I worked on using large transformer models for human motion forecasting. My contributions led to 2x higher model performance and 1.5x reduction in training time and led to a submission to NeurIPS 2021 (Arxiv version will be out soon).

I also worked on one of the most important problems for autonomous vehicles: how do we know if our vehicle is driving well? The industry does not have a clear, widely accepted metric at the moment, yet, human passengers can easily tell if an autonomous vehicle is driving poorly. We try to capture the human intuition with deep neural nets to evaluate the quality of vehicle trajectory.

Select Research Projects:

My Publications:

7 Journal and 10 Conference publications. My Google Scholar Page says I have 298 citations since 2015 and and h-index of 9.

REFERREED JOURNALS:

1. Hao-Tien (Lewis) Chiang, John E. G. Baxter, Mohammad R. Yousefi, Satomi Sugaya, Aleksandra Faust and Lydia Tapia, “Fast Deep Swept Volume Estimators,” International Journal of Robotics Research (IJRR), 2020.

2. Anthony Francis, Aleksandra Faust, Hao-Tien (Lewis) Chiang, Jasmine Hsu, J. Chase Kew, Marek Fiser and Tsang-Wei Edward Lee, “Long-Range Indoor Navigation with PRM-RL”, IEEE Transactions on Robotics (T-RO), 4(36), 1115-1134, 2020.

3. Hao-Tien (Lewis) Chiang, Jasmine Hsu, Marek Fiser, Lydia Tapia and Aleksandra Faust, “RL-RRT: Kinodynamic Motion Planning via Learning Reachability Estimators from RL Policies,” IEEE Robotics and Automation Letters (RA-L), 4(4), 4298-4305, 2019.

4. Hao-Tien (Lewis) Chiang, Aleksandra Faust, Marek Fiser and Anthony Francis, “Learning Navigation Behaviors End-to-End with AutoRL,” IEEE Robotics and Automation Letters (RA-L), 4(2), 2007-2014, 2019. Most read article of RA-L in 2019. Featured in the Google AI Blog, IEEE Spectrum, VentureBeat.com, and PacktPub.com.

5. Hao-Tien (Lewis) Chiang and Lydia Tapia, “COLREG-RRT: A RRT-based COLREGSCompliant Motion Planner for Surface Vehicle Navigation,” IEEE Robotics and Automation Letters (RA-L), 3(3), 2024-2031, 2018.

6. Nicholas Malone, Hao-Tien (Lewis) Chiang, Kendra Lesser, Meeko Oishi, and Lydia Tapia, “Hybrid Dynamic Moving Obstacle Avoidance Using a Stochastic Reachable Set-Based Potential Field,” IEEE Transactions on Robotics, 33(5), 1124–1138, 2017.

7. Hao-Tien (Lewis) Chiang, Guanglei Xu and Rolando Somma, “Improved bounds for eigenpath traversal” Physical Review A, 89, No. 1(012314), 2014.

REFERREED CONFERENCES:

1. Arpit Garg, Hao-Tien (Lewis) Chiang, Satomi Sugaya, Aleksandra Faust and Lydia Tapia, “Comparison of Deep Reinforcement Learning Policies to Formal Methods for Moving Obstacle Avoidance,” In Proceedings of the IEEE International Conference on Intelligent Robots and Systems (IROS), To Appear, 2019.

2. Hao-Tien (Lewis) Chiang, Aleksandra Faust, Satomi Sugaya and Lydia Tapia.“Fast Swept Volume Estimation with Deep Learning.” In International Workshop on Algorithmic Foundations of Robotics (WAFR), Merida, Mexico, 2018 Algorithmic Foundations of Robotics XIII, in press, Springer, 2018. Invited to submit to a special issue of IJRR. Featured in the PC Magazine.

3. Hao-Tien (Lewis) Chiang, Baisravan HomChauhudri, Lee Smith and Lydia Tapia.“Safety, Challenges, and Performance of Motion Planners in Dynamic Environments.” Proceedings of the International Symposium of Robotics Research (ISRR), pp. 1-16, Puerto Varas, Chile, 2017.

4. Torin Adamson, Meeko Oishi, Hao-Tien (Lewis) Chiang, Lydia Tapia , “Busy Beeway: A Game for Testing Human-Automation Collaboration for Navigation,” In Proceedings of the ACM SIGGRAPH Motion in Games (MIG), pp. 9:1–9:6, Barcelona, Spain, 2017.

5. Hao-Tien (Lewis) Chiang, Baisravan HomChaudhri, Abraham P. Vinod, Meeko Oishi, Lydia Tapia, “Dynamic Risk Tolerance: Motion Planning by Balancing Short-Term and LongTerm Stochastic Dynamic Predictions,” In Proceedings of IEEE International Conference on Robotics and Automation (ICRA), pp. 3762–3769, Singapore, May 2017.

6. Hao-Tien (Lewis) Chiang, Nathanael Rackley, Lydia Tapia, “Runtime SES Planning: Online Motion Planning in Environments with Stochastic Dynamics and Uncertainty,” In Proceedings of the IEEE International Conference on Intelligent Robots and Systems (IROS), pp. 4802– 4809, Daejeon, South Korea, October 2016.

7. Aleksandra Faust, Hao-Tien (Lewis) Chiang, Nathanael Rackley, Lydia Tapia, “Avoiding Moving Obstacles with Stochastic Hybrid Dynamics using PEARL: PrEference Appraisal Reinforcement Learning,” In Proceedings of IEEE International Conference on Robotics and Automation (ICRA), pp. 484–490, Stockholm, Sweden, May 2016.

8. Hao-Tien (Lewis) Chiang, Nathanael Rackley, Lydia Tapia, “Stochastic Ensemble Simulation Motion Planning in Stochastic Dynamic Environments,” In Proceedings of the IEEE International Conference on Intelligent Robots and Systems (IROS), pp. 3836–3843, Hamburg, Germany, September 2015.

9. Hao-Tien (Lewis) Chiang, Nicholas Malone, Kendra Lesser, Meeko Oishi, Lydia Tapia, “PathGuided Artificial Potential Fields with Stochastic Reachable Sets for Motion Planning in Highly Dynamic Environments,” In Proceedings of IEEE International Conference on Robotics and Automation (ICRA), pp.2347–2354, Seattle, Washington, May 2015.

10. Hao-Tien (Lewis) Chiang, Nicholas Malone, Kendra Lesser, Meeko Oishi, Lydia Tapia, “Aggressive Moving Obstacle Avoidance Using a Stochastic Reachable Set Based Potential Field,” In International Workshop on Algorithmic Foundations of Robotics (WAFR), Istanbul, Turkey, Aug 2014. Published in H. Akin et al., editors, Algorithmic Foundations of Robotics XI, pp. 73–90, Zeist, Springer, 2015.

PRE-PRINTS

1. Aleksandra Faust, Hao-Tien (Lewis) Chiang and Lydia Tapia.“PEARL: PrEference Appraisal Reinforcement Learning for Motion Planning.” In arXiv preprint arXiv:1811.12651 (2018).