Caterpillar Inc.
Caterpillar Inc.
Caterpillar Inc. je globalno proizvodno podjetje, ki razvija težko opremo in motorje za gradbeništvo, rudarstvo, energetiko ter druge panoge, usmerjene v infrastrukturo. Njegov portfelj izdelkov vključuje bagre, buldožerje, zgibne prekucnike in energetske sisteme, ki jih podpirajo napredna tehnologija ter svetovna mreža trgovcev. Podjetje, ustanovljeno leta 1925, združuje dolgoletno industrijsko strokovno znanje z osredotočenostjo na izboljšanje produktivnosti ter razvoj bolj trajnostne opreme in delovanja.

Prediction Engineer, Autonomous Mining Vehicles

Develop behavior and trajectory prediction for Caterpillar’s autonomous mining vehicles, supporting safer operations and more efficient transport. Build and integrate prediction systems with ROS 2 and C++ for mining environments.

Opis

  • Develop behavior and map-constrained trajectory prediction algorithms for autonomous mining, providing reliable inputs to planning and improving vehicle safety and transport efficiency.
  • Define prediction architecture, inputs, outputs, data flows, and interface contracts with perception, tracking, HD maps, planning, and safety systems.
  • Build workflows for object state modeling, trajectory history, intent recognition, multimodal prediction, confidence reporting, and anomaly diagnosis.
  • Design interpretable, testable prediction frameworks that can be extended to intersections, haul roads, loading and unloading areas, queues, weak networks, and incomplete dispatch information.
  • Model behavior and motion for autonomous and manually operated mining trucks, support vehicles, excavators, loaders, water trucks, and pedestrians.
  • Develop intent recognition and trajectory prediction for following, passing, merging, yielding, intersections, and vehicle exchanges and queues at loading and unloading areas.
  • Use HD maps, road topology, worksite semantics, right-of-way, and dispatch information to generate candidate routes, time occupancy, conflict probabilities, and uncertainty estimates.
  • Work with perception, prediction, planning, safety, and simulation teams so outputs support collision checks, yielding, following, obstacle avoidance, and speed planning.
  • Establish scenario mining, labeling and metrics, simulation replay, issue reproduction, and vehicle validation to expand complex-scene coverage and reliability.
  • Improve prediction latency, resource use, stability, and deployment, including ROS 2 and C++ integration, logging, and troubleshooting.

Zahteve

  • Master’s degree or higher in autonomous driving, robotics, artificial intelligence, computer science, control engineering, vehicle engineering, or a related field.
  • At least five years of algorithm development experience in autonomous driving, robotics, or unmanned systems.
  • Project experience in prediction, tracking, behavior understanding, or planning.
  • Experience building an algorithm module from scratch, defining interfaces and evaluation metrics, and driving ongoing improvement.
  • Familiarity with motion prediction and state estimation methods such as CV, CA, CTRV, CTRA, Kalman filters, EKF, UKF, and IMM.
  • Experience with deep learning prediction methods such as LSTM, GRU, Transformers, GNN, LaneGCN, VectorNet, HiVT, MTR, or Scene Transformer is preferred.
  • Experience in multimodal trajectory prediction, behavior and intent recognition, interaction modeling, trajectory scoring, uncertainty modeling, and confidence calibration.
  • Understanding of HD maps, road topology, map matching, map semantics, route association, and map-constrained candidate trajectory generation.
  • Understanding of how prediction connects to planning, and ability to turn predictions into engineering inputs for safety distances, conflict detection, yielding, following, and obstacle avoidance.
  • Experience with autonomous driving in mines, ports, industrial campuses, logistics yards, construction equipment, or other complex, closed environments is a plus.
  • Experience with mixed traffic, mixed fleets, multi-vehicle coordination, V2X, cloud dispatch coordination, or interaction prediction for non-cooperative vehicles is a plus.
  • Experience with data feedback loops, automated labeling, simulation evaluation, vehicle replay, model deployment, and production issue diagnosis is a plus.
  • Familiarity with the Prediction modules in Apollo, Autoware, or large-scale autonomous driving platforms is preferred.
  • Experience predicting behavior and trajectories on unstructured roads or when map information is incomplete or locally unavailable is preferred.

Ugodnosti

  • Equal opportunity employer.
  • Applicants of all ages who meet the qualifications are encouraged to apply.
  • Opportunity to join the talent community.

Sorodna delovna mesta