Locate named objects using spatial memory and active search.
npx skills add https://github.com/bensonlee5/openral --skill find-object
Please help me install this Agent Skill. Skill: find-object Source: https://github.com/bensonlee5/openral/tree/main/rskills/find-object Command: npx skills add https://github.com/bensonlee5/openral --skill find-object
This Skill helps in locating a named object that is not given a pose and is not currently in view, providing a decision procedure that falls back to active search and human handoff if necessary.
Execute the find-object skill to locate a specific object, e.g., 'Find the mug.'
High-intent search queries and answers about installing and using this skill.
To locate an object not in view, this Skill queries spatial memory to recall its last known location. It then uses active search and navigation to physically find the target item if the initial memory check fails.
If an active search fails to locate the named object within its bounded search budget, the Skill escalates the task through human handoff, prompting human intervention to resolve the missing item.
Yes, locating a named object requires a camera for vision and the physical ability to navigate to and open containers. These capabilities are necessary to check occluding locations during active search.
Robotic task planning handles unknown object poses by using a decision procedure that checks spatial memory first. It then executes a bounded active search across ranked candidate rooms and containers before falling back to human handoff.
The primary limitation of active search is the bounded search budget. If the target object is not found within the allocated search limits across ranked candidate locations, the system must escalate to human handoff.
Yes, you can execute the find-object Skill to locate specific items such as a mug in a kitchen. It uses spatial memory, active search, and container navigation to find the target.
specs-devsProvide real-time per-pixel depth frames from Spectacles via DepthModule.
tcytanGenerate end-to-end implementation plans with exploration artifacts and task files.
zjunlpMove an object from inventory to a specified container in a target scienceWorld room.
KJdotIOConfigure and calibrate multi-sensor perception pipelines in ROS2 robotic systems.
