Robotics Ph.D. candidate ◆ University of Toronto Made in NYC

Jerry (Qǐlóng)
Chéng

程启龙

Robots for the home, built end to end.

Email GitHub X CV Scholar
Jerry (Qilong) Cheng

I’m Jerry (Qilong) Cheng, a Robotics Ph.D. Candidate at the University of Toronto trying to solve Robotics and Embodied intelligence. I work with Prof. Jonathan Kelly in the STARS Lab (UTIAS), and began my Ph.D. at New York University with Prof. Ludovic Righetti (Machines in Motion).

I want to build generalized robots for the homes, automate daily tasks, save us time and enrich our lives.

Research Vision

“We are born too late to explore the Earth and too early to explore the stars — but just in time to solve AI and robotics.” — Jim Fan

Robotics is nearing the scaling inflection that reshaped language and vision — and the binding constraint is large-scale, diverse action data, not new methods. I build one full-stack pipeline aimed at that bottleneck, connecting hardware co-design, learning from human video, and real-world reinforcement learning — toward general-purpose robots for the home.

“Those who are truly serious about software should design their own hardware.” — Steve Jobs

Publications

All publications
Fig. 1 — Evolution of humanoid locomotion control across classical, learning-based, and emerging eras Science Robotics 2026 Evolution of Humanoid Locomotion Control Y. Gu, G. Shi, F. Shi, I. Chang, Y. Wang, Q. Cheng, Z. Olkin, I. Lopez-Sanchez, Y. Feng, J. Zhang, A. D. Ames, H. Su, and K. Sreenath Quick look IRIS cinema robot arm CRV (Oral, top 10%) 2026 IRIS: Learning-Driven Task-Specific Cinema Robot Arm for Visuomotor Motion Control Q. Cheng, M. Mackay and A. Bereyhi Quick look A camera orbiting an AprilTag target: the AX = YB kinematic loop IJRR 2025 Certifiably correct hand-eye calibration E. Wise, P. Kaveti, Q. Cheng, W. Wang, H. Singh, J. Kelly, D. M. Rosen, and M. Giamou Quick look IRIS cinema robot arm thesis cover M.Eng. Thesis 2025 IRIS: A Low-Cost, 3D-Printed Cinema Robot Arm (M.Eng. Thesis) Q. Cheng, M. Mackay, A. Bereyhi Quick look VibraForge toolkit CHI 2025 VibraForge: A Scalable Prototyping Toolkit For Creating Spatialized Vibrotactile Feedback Systems B. Huang, S. Ren, Y. Luo, Q. Cheng, H. Cai, Y. Sang, M. Sousa, P. H. Dietz, and D. Wigdor Quick look AeroHaptix UAV teleoperation IEEE RA-L 2024 AeroHaptix: A Wearable Vibrotactile Feedback System for Enhancing Collision Avoidance in UAV Teleoperation B. Huang, Z. Wang, Q. Cheng, S. Ren, H. Cai, A. Alvarez Valdivia, K. Mahadevan, and D. Wigdor Quick look Stair scenes: RGB, estimated normals, extracted tread planes TRC Poster 2024 BipedNav: stair geometry from one image, then walked A. Qiu, A. Narayanan, Q. Cheng, and B. Laschowski Quick look Two radars on a vehicle with Doppler returns and ego-velocity arrows IEEE AIM 2023 Radar-to-radar calibration from ego-velocity Q. Cheng, E. Wise, and J. Kelly Quick look Handheld radar-camera rig flying along a continuous-time B-spline IEEE TRO 2022 Radar-camera spatiotemporal calibration E. Wise, Q. Cheng, and J. Kelly Quick look Weakly supervised augmentation overview IEEE Access 2022 Weakly Supervised Semantic and Attentive Data Mixing Augmentation for Fine-Grained Visual Categorization M. He, Q. Cheng, and G. Qi Quick look Generative design for a unicycle robot SPIE ACAIB 2022 Generative Design for Self-Balancing Unicycle Robot in Additive Manufacturing J. Chen, Q. Cheng, and M. Han Quick look

Projects

All projects
Two mobile manipulators carrying a shared object in MuJoCo Robot learning 2026 Read the Room: shared policy, shared load Grasp-aware planning and shared learned local control for cooperative transport. Explore the paper, seven-stage task visualization, 35/40 frozen missions across ten scenes, and interactive trajectories. Separate development experiments study recovery and local coordination. Quick look MABEL, an open-source anthropomorphic mobile bimanual robot (paper Figure 1) Open hardware 2026 MABEL: Open Mobile Bimanual Robot Open-source (MIT) whole-body robot: a 3-module holonomic swerve base, a 0.635 m lift with actuated torso, two 7-DOF arms, two 17-DOF ORCA hands, and a 3-DOF active head. 56 DOF from $8,722, with no machine shop. Teleoperated from Apple Vision Pro, iPhone, browser, or gamepad; full MuJoCo digital twin; demonstration-to-policy pipeline (ACT, Diffusion Policy, π0) deployed onboard. System paper under review. Quick look GeoDEX dexterous manipulation Robot learning 2026 GeoDEX: a human hand onto 23 robot joints Apple Vision Pro demonstrations retargeted onto a 23-DoF xArm + ORCA hand by a bifurcated SE(3) optimisation, then cloned by a chunked transformer with a grasp-weighted loss. State Bottlenecking — a mandatory home pose between sub-tasks — takes two-object clearing from 22 % to 88 %. The page lets you switch each term of the retargeting cost off and watch the pinch geometry come apart. Quick look TINTIN lunar landing RL Reinforcement learning 2025 TINTIN: fly the lunar lander yourself 6-DoF variable-mass lunar lander in MuJoCo with a 2-DoF gimbaled engine. A SAC policy learns guidance and control as one map; a cascaded PD baseline, re-run for 500 descents, lands 70.2 % of the time and comes apart above 50 m/s entry speed. The page lets you fly the lander and re-run the Monte Carlo in your browser. Quick look Stair point-cloud profile with fitted tread and riser planes Perception 2024 Monocular Depth Estimation for Human-Robot Locomotion Monocular depth pipeline for human-robot walking environments: Metric3D v2 fine-tuned on 16,780 RGB-D stair images, metric reprojection to 3D point clouds, and plane extraction for stair rise and run. Re-run in 2026 on the archived scenes: 1.7 cm rise / 2.5 cm run mean error against a tape measure. Quick look CLEAR-Net cart-pole reinforcement learning Reinforcement learning 2024 CLEAR-Net: train a cart-pole in your browser Tabular Q-learning, DQN, PPO and SAC on a cart-pole whose every observation is corrupted by Gaussian noise. A five-seed re-run found the released bin limits collapse cart position into two of eight bins and double the drift, that survival peaks at σ = 0.1 rather than at zero noise, and that a coarse grid beats a fine one fifteen-fold at a fixed budget. The whole tabular experiment runs live in the page. Quick look Five-link biped with knees walking under virtual holonomic constraints Perception + Control 2024 BipedNav: see the stairs once, then walk Monocular stair/obstacle geometry from one image feeds a five-link biped with knees walking by virtual holonomic constraints (hybrid limit cycles, verified against their return maps). From one view of an obstacle the planner schedules the approach and two step-over gaits chained through the impact map. Includes the re-test of the two-link ECE1658 design and an in-browser demo. Quick look Ground truth, dead reckoning and two estimates among landmarks State estimation 2023 Where am I? Batch vs sliding window Maximum-a-posteriori estimation over every pose at once, on SE(3), with a stereo camera and an IMU — and the sliding window that approximates it. A four-seed re-run found batch is both more accurate and cheaper than a wide window over a whole trajectory, and that a window looks over-confident only when its noise model is wrong. Drag the window size in the page and watch the estimate move. Quick look Arduino mobile robot Mechatronics 2022 Four ultrasonics, no compass, one bit rotation A 170 mm robot finds itself in an 8 × 4 maze using four ultrasonics and nothing else: a tile is four bits of wall, and because the bits run around the tile, turning 90° is a bit rotation. Localization becomes a set intersection over 24 tiles × 4 headings. Swept exhaustively, no single reading ever localizes it — but with the gyro trusted, every start resolves in one tile-move. Drive it on the page and watch the candidates disappear. Quick look 8x8x8 LED cube Embedded 2022 512 lights, nine chips, one illusion 512 hand-soldered LEDs driven by nine daisy-chained shift registers over one SPI bus. Only one of the eight layers is ever lit; the cube is an illusion held together by refreshing faster than the eye can follow. Twelve animations and a shallow-water liquid you tip by dragging — volume held to 10−13, surface level to 0.008 cells per column — plus 3-D Snake with wrap-around walls and Pong against a machine that solves for where the ball will cross its own face. Drag the refresh rate down and watch persistence of vision give out. Quick look Turtlebot2 SLAM and navigation SLAM 2022 Where am I? Exploring an unknown room A TurtleBot 2 mapping an unknown 6×6 m room with a Kinect, three bumpers and GMapping, switching between reactive, behaviour and deliberate control. A six-room re-run found the four strategies are separated by less than their own variance, and that the hybrid controller is simultaneously the best and the worst of them. Watch the occupancy grid fill in live, and race all four. Quick look Formula racing car drivetrain Mechanical design 2022 Four times the torque, 176 mm of wheelbase back Custom 4:1 planetary reduction for the UofT Formula team's first electric car, quadrupling torque at an estimated 91 % efficiency. Re-checking the design against its own numbers: the gearbox cuts the motor-to-differential span from 273 mm to 96 mm, two of the four shortlisted tooth sets cannot take four equally spaced planets, and a rad/s-labelled-Hz slip put the system's natural frequency 2π too high. Spin the gear set on the page. Quick look

News

University of Toronto campus, University College
STARS Laboratory logo

[09/2026] Transferring my Ph.D. to the University of Toronto to work with Prof. Jonathan Kelly at the STARS Laboratory (UTIAS) — building dexterous manipulation on the road to solving robotics.

Figure 1 — three eras of humanoid locomotion control: classical, learning-based, emerging
Classical control versus reinforcement learning pipelines
Emerging shifts in humanoid locomotion control

[08/2026] Our review “Evolution of Humanoid Locomotion Control” was published in Science Robotics (Vol. 11, Issue 117) — from classical control through learning to foundation and world models. Reading list on GitHub.

Oral presentation of IRIS at CRV 2026
AI/CRV 2026 conference badge — Jerry Cheng, New York University
Simon Fraser University campus, Vancouver

[05/2026] Our paper IRIS was accepted to CRV 2026 (Conference on Robots and Vision) — I attended and gave an oral presentation at Simon Fraser University, Vancouver.

[11/2025] Graduated with a Master of Engineering (MEng) in Electrical & Computer Engineering (ECE) at the University of Toronto.

[10/2025] Attended IROS 2025 in Hangzhou, China and presented “AeroHaptix”.

Machines in Motion Lab group, NYU
New York City skyline
Jerry with a humanoid robot at NYU

[09/2025] Started my Ph.D. in Mechanical Engineering at New York University, joining Prof. Ludovic Righetti's Machines in Motion Lab — working with Ludovic and Katsuo.

Robot-world & hand-eye calibration figure — IJRR 2025

[07/2025] Our paper on certifiably correct robot–world & hand-eye calibration was accepted to the International Journal of Robotics Research (IJRR).

[04/2025] Attended CHI 2025 in Yokohama, Japan and presented “VibraForge”.

[11/2024] Presented at ARIA 2024 in Toronto with our VibraForge toolkits.

[07/2024] Presented monocular depth estimation work at the Toronto Robotics Conference 2024.

[11/2023] Demonstrated “VibraForge” at ARIA 2023 in Toronto.

[06/2023] Graduated from the University of Toronto with a Bachelor of Applied Science (BASc) in Mechanical Engineering — mechatronics & robotics focus.

Coffee Chat

I'm always open to chatting about Robotics, Research, Startups, or just connecting! Feel free to grab a time on my calendar below.

Location: Any Cafe with Enough Clearance for a 6-DOF Robot Arm

Schedule a Coffee
Total Visitor Analytics