[Job Order ID: 976422]

Responsibilities

Trajectory Prediction Models: Develop advanced learning-based models for predicting the movement of vehicles, pedestrians, and cyclists using sophisticated architectures like transformers, diffusion models, or graph neural networks.

Multi-Agent Interaction Modeling: Integrate end-to-end neural networks that can jointly predict the motion of multiple agents, enabling the modeling of complex interactions in dynamic environments.

Uncertainty-Aware Forecasting: Enhance probabilistic prediction with self-supervised learning techniques, focusing on robustness in scenarios with low-frequency events or long-tail distributions.

Social and Cooperative Behavior Modeling: Apply game-theoretic and reinforcement learning approaches to better simulate social and cooperative behaviors in autonomous systems.

Reinforcement and Imitation Learning for Decision-Making: Design decision-making systems using reinforcement learning (RL) or imitation learning (IL) that adapt to complex real-world driving situations and handle uncertainty effectively.

Multi-Agent Decision-Making: Implement multi-agent decision frameworks that allow autonomous vehicles to interact and navigate through dense, uncertain, and highly interactive environments.

Hybrid Motion Planning Systems: Develop motion planning systems combining classical planning methods (graph search, optimization) and learning-based approaches (RL, IL) for improved safety, efficiency, and comfort.

Differentiable Motion Planning: Explore differentiable motion planning architectures that allow the backpropagation of gradients through neural networks, improving the optimization of real-time trajectory generation.

Closed-Loop Simulation & Self-Supervised Learning: Create closed-loop simulation frameworks for seamless integration of prediction, decision, and planning, and apply multi-modal self-supervised learning techniques to improve system robustness.

Real-World Deployment and System Performance: Ensure the real-time deployment of models on embedded platforms with techniques like model quantization and pruning; also implement continuous performance improvement through online learning and adaptation from fleet data.

Requirements:

Strong knowledge of behavior prediction, decision-making, and motion planning for autonomous systems.

Proficiency in deep learning frameworks (PyTorch/TensorFlow) with experience in transformers, graph neural networks, diffusion models, and reinforcement learning.

Experience with planning and control algorithms (A*, RRT, ILQR, MPC, PID, model-free RL).

Strong programming skills in C++ and Python, with experience in high-performance computing and real-time embedded system optimization.

Familiarity with robotics middleware (ROS/ROS2, Apollo, Autoware) and real-time embedded systems.

Understanding of probabilistic decision-making, POMDPs, and Bayesian inference for uncertainty-aware driving strategies.

Experience in end-to-end autonomous driving, differentiable planning, or foundation models for driving.

Publications in top conferences (CVPR, NeurIPS, ICRA, CoRL, ICCV) related to behavior prediction, RL-based decision-making, or motion planning.

Background in game theory, multi-agent reinforcement learning (MARL), or traffic simulation-based planning.

To Apply, please kindly email your updated resume to cv_gary@goodjobcreations.com.sg