6.8210 Spring 2023 Final Project Presentation (Audio fixed)

6.8210 Spring 2023 Final Project Presentation (Audio fixed)

by underactuated

Applied Sciences & Engineering Control & Robotics TJFMAutomatic control engineeringTJFM1Robotics
🎙 underactuated 👥 17K 📅 May 18, 2023 ⏱ 110 min 👁 1K 📄 original study 🧭 2026-08-05
Available in: English (current) Français

Keywords

GCSdiffusion policytrajectory optimizationperchinghumanoid

Summary

This video is a recording of the final project presentations for MIT’s 6.8210 course (Spring 2023), focusing on underactuated robotics. It features three student projects. The first project uses Graph of Convex Sets (GCS) to generate demonstrations for a diffusion policy to push a box with a point finger, achieving a 77% success rate. The second project optimizes a parkour vault trajectory for the MIT humanoid using direct collocation, reducing torque requirements. The third project develops a perching controller for a flapping micro aerial vehicle (RoboBee) using trajectory optimization and LQR. The presentations include technical details, results, and Q&A sessions. The video is aimed at an academic audience and provides insights into advanced robotics research.

115 words

Critical Evaluation

The video presents three distinct robotics research projects, each with a clear problem statement, methodology, and results. The first project on GCS and diffusion policy is particularly well-structured, demonstrating a novel approach to combining motion planning with learning-based policies. The use of GCS to generate demonstrations addresses the limitations of human demonstrations, and the diffusion policy shows robustness to disturbances. However, the success rate of 77% indicates room for improvement, and the authors acknowledge limitations with out-of-distribution states. The second project on humanoid parkour is also impressive, using direct collocation to optimize a complex dynamic maneuver. The cost function tuning is discussed, showing a systematic approach to improving trajectory quality. The third project on RoboBee perching is more preliminary, with simulations only, but it lays groundwork for future hardware implementation. Overall, the presentations are technically rigorous, but they lack formal citations and peer review. The Q&A sessions add value by clarifying design choices and limitations. The video is suitable for an advanced audience familiar with robotics and control theory.

169 words

Title / Content Match

The title accurately describes the content: it is a final project presentation for a course, and the audio has been fixed.

Quality & Reliability

7/10

The video presents final project presentations from a graduate robotics course, showcasing original research projects. The content is technically detailed and appears rigorous, but it is not peer-reviewed and lacks formal citations. The presentations are from students, so the authority is moderate.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Contribution & Novelties

The video showcases novel applications of existing techniques: using GCS for motion planning with contact, combining it with diffusion policies, and applying trajectory optimization to humanoid parkour and micro aerial vehicle perching. The projects demonstrate innovative integration of methods.

Pour aller plus loin :

  • Graph of Convex Sets — Background on convex optimization relevant to GCS.
  • Diffusion Policy — Official project page for diffusion policy.
  • Direct Collocation — Overview of direct collocation method used in trajectory optimization.

77 words

Radar Profile

The radar profile shows high scores in quantity and technical level, indicating a dense, technical presentation. Quality and reliability are moderate, reflecting the lack of formal citations and peer review.

Reliability 6/10