Skip to content
Drix10 Blog

Direct robot taught to move in 15 minutes

, 8 items in AI and Robotics Applications, 4 min read

In this digest (8 items)

Direct robot learned to move around in only 15 minutes. The robot used that mobility data to complete a task that requires both mobility and manipulation.

Key points

  • Taught robot to move around in 15 minutes

  • Mobility data enabled task requiring mobility and manipulation

Sources

ParallACT benchmark shows 16% success for Opus 4.8 on multi-agent tasks

ParallACT is a benchmark of multi-principal multi-agent tasks requiring coordination. Frontier models struggle, achieving only 16% success for Opus 4.8 even when they can talk.

Key points

  • Benchmark: ParallACT evaluates coordination in large-repo, multi-agent scenarios.

  • Result: Opus 4.8 reaches 16% success on the benchmark.

Sources

PanoTac dense three-axis tactile sensor demo tube insertion and stopper wiggle

PanoTac is a dense three-axis tactile sensor used in a demo where a robot reorients and inserts a tube, then wiggles the stopper free. The right-hand view shows tactile signals that let the robot detect when the stopper releases and when to change from wiggling to lifting.

Key points

  • Demo: robot reorients and inserts a tube, then wiggles the stopper free while the tube sits loosely in the rack.

  • Right side shows tactile signals that help detect stopper release and when to switch from wiggling to lifting.

Sources

3D population pyramid by Marc Thévenin – age distribution

3D population pyramid by Marc Thévenin of @InedFr visualizes age distribution for a given year and cohort survival across years. It also displays the number of young people across years.

Key points

  • Age distribution in a given year visualized in 3D.

  • Cohort survival and young-people count shown across years.

Sources

OpenDubs robot autonomous 6-hour Halloween candy shift at UW CSE

OpenDubs robot worked a 6 hour shift autonomously handing out candy to trick-or-treaters for Halloween at UW CSE. The project used a full-stack approach.

Key points

  • Robot: OpenDubs

  • Shift length: 6 hours autonomous operation

Sources

DVA robot learns mobility and manipulation in 15 minutes

DVA robot learns mobility and manipulation in 15 minutes. The system uses a single DVA with joint prediction of arm movement and robot maneuvering, without a separate mobility or navigation stack.

Key points

  • No separate mobility or navigation stack; uses joint prediction of arms and robot movement.

  • Mobility data required: only 15 minutes.

Sources

RoboColosseum announced, models to be tested on robots in 48 hours

RoboColosseum is announced as a new project that will put AI models on real robots. A fleet of robots is assembled and the tests will go live in 48 hours.

Key points

  • Project name: RoboColosseum

  • Test start time: coming online in 48 hours

Sources