Thomas Lew
@thomas__lew
Followers
220
Following
147
Media
8
Statuses
31
Research Scientist at @ToyotaResearch. Optimal Control, Machine Learning, Robotics. PhD @Stanford. Previously intern at @Google, @NASAJPL.
Joined March 2020
TRI's latest Large Behavior Model (LBM) paper landed on arxiv last night! Check out our project website: https://t.co/n0qmDRivRH One of our main goals for this paper was to put out a very careful and thorough study on the topic to help people understand the state of the
8
107
489
4/4) This work combines known pathwise approaches to SDEs from the mathematics & machine learning communities and ideas from indirect optimal control. We use proof techniques developed by Pontryagin and recent integrability results on rough paths (Cass, Friz, et al. ~2013).
0
0
0
3/4) As an application, we propose the first indirect shooting method for nonlinear stochastic optimal control: It converges much faster than a direct method on a regulation task. Why? Because it only searches over the initial conditions of the adjoint vector 💡
1
0
0
2/4) Our new optimality conditions are derived via rough path theory instead of Ito calculus. They apply to systems following SDEs driven by Brownian motion, yet they do not rely on FBSDEs. The adjoint equation is interpreted *pathwise* like in the deterministic setting.
1
0
0
1/4) Previous results use forward-backward SDEs (FBSDEs), which can be difficult to use in algorithms. In the deterministic case, optimality conditions inform efficient methods leveraging the structure of solutions. Can we also derive simple conditions for the stochastic case?
1
0
0
I'm excited to share new optimality conditions for nonlinear stochastic optimal control, and the first indirect shooting method for solving these problems! 📖 https://t.co/JdG5NSjQou 💡 How? Using rough path theory ⬇️
2
3
20
Our team at TRI is hiring a research intern for the summer of 2025! An exciting opportunity to pursue research at the intersection of perception and control, and to deploy models and algorithms on high-performance cars https://t.co/rcPpHvPYHG
0
3
14
Considering or just starting a PhD in robotics, optimization, ML, or control? If so, you may find this interview interesting: https://t.co/5vmmWNXpCW Thanks again to Rodolphe Sepulchre for the invitation! @IEEEorg @CSSIEEE
0
6
17
Presenting our risk-averse trajectory optimization work at #ICRA2024 on Wednesday at 2.45pm in AX-206. Come have a chat about planning under uncertainty and optimization! 📖
0
0
5
Our team at TRI is hiring a research intern for this summer! An exciting opportunity to pursue research on machine learning and control and deploy models and algorithms on high-performance cars
2
4
42
@IEEECDC2023 @drmapavone Thank you to all my amazing collaborators!
0
0
2
@IEEECDC2023 @drmapavone and the Control Systems Magazine Outstanding Paper Award for "Convex Optimization for Trajectory Generation: A Tutorial on Generating Dynamically Feasible Trajectories Reliably and Efficiently" https://t.co/j60OT25KpS with R. Bonalli, @drmapavone, and collaborators at UW
1
0
4
The @IEEECDC2023 Outstanding Student Paper Award for "Exact Characterization of the Convex Hulls of Reachable Sets" https://t.co/00nCIdJw8F with Riccardo Bonalli and @drmapavone
1
0
3
I'm excited to share that we just received two outstanding paper awards at @IEEECDC2023 !
3
0
26
I am grateful to everyone who supported me over the years and look forward to working with a great team at TRI!
1
0
1
I am excited to share that I joined the Toyota Research Institute as a research scientist! Our team is developing next-generation assistive driving tech at the handling limits
cnet.com
This technology could automatically give you the skills of a professional driver to help maintain control in at-the-limit situations.
4
0
35
📢 Announcing the first @corl_conf workshop on Out-of-Distribution Generalization in Robotics: Towards Reliable Learning-based Autonomy! #CoRL2023 🎯 How can we build reliable robotic autonomy for the real world? 📅 Short papers due 10/6/23 🌐 https://t.co/PulDqjtm5d 🧵(1/4)
sites.google.com
Workshop Overview
1
15
33
Presenting our robotic table wiping work today at 3pm, Pod 15 at #ICRA2023 ➡️ Come have a chat about reinforcement learning, trajectory optimization, and stochastic dynamics modeling
📢Excited to share our #ICRA2023 work on robotic table wiping via RL + optimal control! 📖 https://t.co/FoS0AioyNA 🎥 https://t.co/9RCsHBzv17 💡RL (for high-level planning) + trajectory optimization (for precise control) can solve complex tasks without on-robot data collection ⬇️
0
2
14
Read how we enabled a robot to reliably wipe up crumbs and spills with an approach for robotics applications in complex environments that uses an #RL policy (trained with a stochastic differential equation simulator) followed by a trajectory optimizer. → https://t.co/Iw7pjVBjac
22
80
309