Multi-modal algorithm (force/touch) engineer

You can sign an employment contract with the Hong Kong company entity, but you must accept full-time office in Shenzhen!!!!

Duties:

1. Force/tactile data processing and modeling:
i. Develop high-precision six-axis force sensor signal processing algorithm, solve the problem of inter-axis coupling, noise suppression and nonlinear correction.
ii. Design a tactile sensor dynamic response model, optimize the interpolation algorithm to eliminate insensitive regions, and solve the problems of lag and drift.
2. Multimodal perception fusion:
i. Integrate visual, haptic, and tactile data modalities to develop deep learning-based environment interaction perception algorithms (such as material recognition, soft object grasping).
ii. Combine bio-inspired methods (such as the human skin tactile mechanism) with sensor layout and data fusion strategy optimization.
3. Force feedback co-simulation:
i. Develop impedance control, admittance control, etc. force feedback algorithms, realize robot compliant operation (such as assembly, medical surgery scenarios).
ii. Design a multi-sensor fusion motion control framework to improve the dynamic balance and precision of complex contact tasks for robots.
4. Algorithm deployment and optimization:
i. Lead the algorithm migration from simulation environment (MuJoCo/Gazebo) to real robot platform, solving sensor calibration and hardware latency issues.

Qualifications:

1. Academic background Master's degree or above, robotics, computer science, electronic engineering and other related fields, QS Top 200/985/211 is preferred.
2. Proficient in the principles of force/touch sensors, familiar with the data characteristics of mainstream devices.
3. Master the multi-modal fusion framework, and be familiar with PyTorch/TensorFlow and robot middleware (ROS/Isaac Sim).
4. Have reinforcement learning (RL), imitation learning (IL) experience, be familiar with Voxposer, RT-1 and other embodied intelligent models.
5. In CVPR/ICRA/IROS/RSS and other top conferences, those who publish multimodal perception or force control related papers are preferred.

ii. For embedded systems (such as FPGA/ROS nodes), deploy algorithms with lightweighting.