JAMES WU
DESIGN ENGINEER
01 / Physical / Ongoing industrial placement / robotics

Hands-on work with humanoid and wheeled robots.

At Dobot Robotics, I worked with humanoid and wheeled robots on operation, policy-related data collection and hardware assembly. My placement also covered navigation experiments, vision-guided manipulation and speech interaction.

PROJECTDobot Robotics
MY ROLETechnical Support Engineer Intern · robot applications and testing
CONTEXTDobot Robotics
Jun–Sep 2026 · ongoing
Dobot robot performing a depth-camera-guided pick-and-place demonstration with a bottleInspect full-size image ↗
FIG. 01Robot demonstration during the Dobot placement.
01 / OVERVIEW

The project in a minute.

THE OUTCOME

Hands-on experience connecting physical robots, perception, control and services through task demonstrations, teleoperation, navigation experiments and hardware work.

What supports this? ↓
The challenge
Make concrete tasks work on humanoid and wheeled robots, where perception, control, communication and hardware all affect the result.
My contribution
Technical Support Engineer Intern · robot applications and testing. Placement reports document teleoperated manipulation data collection, a calibrated vision-guided pick-and-place demo, navigation experiments and speech-service testing.
Read this in context
Internship contributions across several workstreams. The manipulation result is limited to the calibrated demonstration; exploratory navigation is not a benchmark success.
01Robot operation02Policy-related work03Hardware assembly
02 / THINKING

The choices behind the result.

Open a chapter to follow the work and the choices made along the way.

01 / DESIGN

Collect demonstrations with a consistent task structure

Teleoperated manipulation tasks were divided into clear phases and short action chunks. Environmental variation was added after the task sequence was stable.

Inspect the supporting evidence ↓
02 / ENGINEERING

Use a bounded pipeline for the manipulation demo

The documented workflow combines object detection and segmentation, depth and hand-eye calibration, then robot-frame targets and fixed grasp/transfer waypoints. Results apply to a fixed object in the calibrated workspace.

Inspect the supporting evidence ↓
03 / ENGINEERING

Distinguish navigation experiments from proven performance

I explored planning with an incomplete map in Isaac Sim, using global guidance and local MPPI. The review’s replay reached the step limit; its value is the decision trace and failure analysis, not a successful navigation benchmark.

Inspect the supporting evidence ↓
04 / ENGINEERING

Connect speech to the right source of information

Guide-robot work covered speech recognition, routing between a knowledge base and a general model, speech output and robot actions. Diagnostic work also involved reproducing faults and documenting operating conditions.

Inspect the supporting evidence ↓
METHODS USEDRobot operationHardware assemblyVR teleoperationVision-guided manipulationNavigationSpeech interaction
03 / EVIDENCE

Results, and what they mean.

Internship contributions across several workstreams. The manipulation result is limited to the calibrated demonstration; exploratory navigation is not a benchmark success.

Physical robots

Humanoid and wheeled platforms

Robot operation, policy-related demonstration collection and hardware assembly.

Calibrated demo

Vision-guided manipulation

Repeated transfer of a fixed object within a calibrated workspace. No general-purpose grasp success rate is claimed.

System practice

Navigation, speech and integration

Application development and testing across perception, control, services and hardware.

Source notes & scope 2
  1. Placement Review SUAT / Dobot v9slides 10–15

    High-level workstreams and evaluation boundaries.

  2. James’s description of this placementAuthor clarification

    Humanoid and wheeled robots, policy-related work and hardware assembly.

Dobot humanoid robot with a communication board about speech interaction and diagnostic handoffs↗
FIG. 02Humanoid robot during the Dobot placement.
Dobot robot collecting teleoperated manipulation data with a blue garment↗
FIG. 03Robot operation during the Dobot placement.
↑ Return to the overview