top of page

Mobile Manipulator Development



We are developing mobile manipulators (self-propelled drones with arms) based on our experience and achievements in RoboCup2022 and WRS. After 10 years of hard work, we are finally catching up with Fetch.
The intended task is shopping at convenience stores.This video shows the autonomous navigation between low/high-level areas via relay floor by elevator cooperation.

Concept, Outline and Policy



We aim to develop robots that can act independently without human intervention, and that are inexpensive and easy to develop (based on ROS). The planned implementation functions are as follows :

  • door opening

  • pressing buttons

  • pick/place objects

  • elevator riding

  • autonomous navigation

Expected Task

The "Convenience Store Task" in competitions such as RoboCup and WRS is designed to simulate the movements of the store staff, but the task we aim for is shopping at a convenience store from the user's perspective.
For the robot to make purchases at a convenience store, it must perform the following flow :

  1. User: Speak what to order (voice recognition)

  2. Robot: Heading to the convenience store, opening the door, boarding the elevator, and arriving at the convenience store

  3. Robot: Picks up merchandise (aligns AR markers on shelves,  recognizes and takes out merchandise)

  4. Robot: Places merchandise at the cash register

  5. Clerk: Checkout (input to POS, read PayPay QR code attached to the robot)

  6. Robot: Receives products in a bag

  7. Robot: Moves to the user's location and hands over the goods to the user upon arrival




We are creating a prototype by combining commercial products to develop within a short period quickly. We combined an autonomous base (large enough to pass through the width of a convenience store aisle) with a manipulator to perform convenience store tasks (picking items from the back of the shelves).

  • Name : SOAR

  • Size : (h)1,360mm~1,820mm, (w)550mm, (d)650mm

  • Weight : 91.65kg

  • Operation time : About 45 minutes (actual measured value)

  • Maximum speed : 3.0 km/h (measured value)

  • Sensor configuration : 2D Lidar, IMU, Depth Camera, Microphone, Thermal Camera, Zoom Camera, LED lights

  • Degree of freedom of the arm : 7

Components (base)



  • Large enough to pass through the width of a convenience store aisle.

  • Tires with suspension for smooth ride on elevator.

  • Stable autonomous movement even with an elevating arm attached.

Components (linear actuators)



  • Capable of picking items from the back of convenience store shelves

  • Raise and lowers arm and head at the same time

  • Self-weight compensation mechanism for arm weight (does not descend with arm weight)

Pick & place of objects using image recognition


This is a pick & place video of cup noodles using image recognition.

Development scene

bottom of page