Robotics & Autonomy
A published NVIDIA application — not a granted patent — claims an autonomous machine that localizes road markings by seeding a LiDAR refinement with a camera estimate. Its three independent claims land at the machine, system, and system-on-a-chip level, classified under the G06T image-analysis family.
Robotics & Autonomy
A patent granted June 30 covers the head-and-neck assembly of a humanoid robot with three independent claims — one built around a light-passing, sensor-bearing shell, one adding a tinted low-diopter shell, one around a two-camera arrangement. It is an issued grant, not a pending application, classified under the B25J manipulator family.
Robotics & Autonomy
A Samsung application published June 25, 2026 is directed to a robot that maintains a dynamic scene graph carrying per-object history — what moved, when, and on whose instruction. It is a pending application, not a granted patent, and it lands in a cluster of recent robot-control filings classified under G05D.
Robotics & Autonomy
A patent issued this week to Kawasaki Heavy Industries claims a substrate-transfer arm that places or picks up a wafer without ever stopping the end effector. Claim 1 fences off the teaching-point state switch, not the gripper, and it lands in the B25J 9/1664 program-control neighborhood.
Robotics IP
G05D is the Cooperative Patent Classification subclass for controlling the position and course of vehicles. Here is what the class covers, how G05D 1/00 anchors autonomous navigation, and how a real G05D claim reads.
Robotics IP
An end-effector is the tool at the end of a robot arm — the gripper, hand, or instrument that does the work. Here is what the term means and how end-effector inventions are claimed in patents.
Robotics IP
Motion planning is how an autonomous system decides the path it will take. Here is what path planning and trajectory planning mean in patents and how a real motion-planner claim is written.
Robotics IP
Sensor fusion combines data from multiple sensors into a single, more reliable view of the world. Here is what the term means in self-driving patents and how it appears in real claim language.
Robotics IP
SLAM lets a robot build a map of an unknown space while tracking its own position in it. Here is what the term means and how it shows up in the language of real, granted patent claims.
Robotics IP
B25J is the Cooperative Patent Classification subclass for manipulators and robots. Here is what the class title actually covers, how its subgroups are organized, and how to read a real B25J patent claim.
Robotics
An arXiv preprint studies path planning for continuum robots, modifying genetic and A* algorithms with an Analytical Hierarchy Process step that weighs distance, motor damage, mechanical damage, and accuracy.
Robotics
A preprint posted to arXiv describes generating robot-hand mechanisms directly from large-scale human demonstrations, optimizing tree-structured designs to reproduce fingertip motion rather than learning a controller per candidate design.
Robotics & Autonomy
A Mobileye application published June 18, 2026 claims a crowd-sourced map that learns, junction by junction, which traffic light governs which drivable path — part of a four-application cluster that lands squarely in the camera-and-map (G01C 21/38) navigation classes.
Robotics
Diffusion visuomotor policies running asynchronously stitch together action chunks that don't agree at the seams. LAGO Policy fixes the discontinuities and adds explicit collision-free planning on top.
Robotics
Engagement-aware MPC rethinks human-in-the-loop control: instead of pulling users in only at failures, it plans interaction to keep them involved while respecting a workload limit.
Robotics
An energy-aware distributed framework splits decision-making between a robot and a remote controller, letting the agent choose whether to run wildfire detection onboard or offload it while a forward-looking model anticipates outcomes before committing.
Robotics
A new paper models multi-agent AI systems as message passing on the same sparse graphs that underlie LDPC codes, then extends density-evolution theory to predict what fraction of a task will go unresolved.
Robotics
Instead of treating sensor field-of-view as a heuristic afterthought, FLAP folds FOV geometry into the drone's trajectory optimization, letting a UAV fly aggressive 3D maneuvers in unknown space while still looking where it's about to go.
Robotics
A new radar-SLAM framework pairs a Rao-Blackwellized particle filter with incremental Hilbert-space Gaussian process mapping to build dense, uncertainty-aware maps from measurements too sparse for LiDAR-style methods.
Robotics & Autonomy
U.S. Patent No. 12,656,500, granted to GM Cruise on June 16, 2026, detects adverse weather at the LiDAR sensor level by measuring how the statistics of the point cloud drift, flagging an environmental change before the perception stack ever sees it.
Robotics & Autonomy
U.S. Patent No. 12,656,493, granted to Waymo on June 16, 2026, covers a rotating LiDAR that switches its laser modulation scheme depending on where the beam is pointed — a power-budget trick aimed squarely at seeing farther down the lane that matters.
Robotics/autonomy
ORCA is an open-source stack that puts control, simulation, VR teleoperation, and hand retargeting behind one interface and plugs dexterous anthropomorphic hands into the same robot-learning pipelines used for simple grippers.
Robotics/autonomy
CADET audits, benchmarks, and repairs the causal confusion in already-deployed end-to-end driving planners, the tendency to tie decisions to co-occurring scenery, without retraining the model.
Robotics/autonomy
SLS-squared plans motion directly from camera images using a learned latent world model, then makes the plan provably safe by combining conformal prediction with GPU-accelerated system level synthesis robust MPC.
