Starting date: April the 20th 2019
Duration: 20 months

 
Project category

OPTION 1: develop a benchmarking solution for one specific benchmarking scenario including one or more outcomes: testbed, software routines and/or experimental datasets.

 
Application

Wearable robots

 
Outcome

Testbed, Sofware & Dataset

 
Team:

COORDINATOR:

  • Politecnico di Milano (Italy)

Partner 2:

  • Sapienza Università di Roma (Italy)

Partner 3:

  • Nova Labs s.r.l. (Italy)

Partner 4:

  • IST-ID (Portugal)
Abstract
The Benchmark-Enabling Active Shopping Trolley (BEAST) is a ​benchmarking testbed with data acquisition and mechanical action capabilities. Its key element is an ​active shopping trolley which, while maintaining all the features of a standard trolley, enables quantitative characterisation of the capability of a robot to manoeuver it safely and effectively. The BEAST trolley can measure its own acceleration, the geometric relationship with surrounding obstacles (distance and collisions) and the forces applied by the robot to its handle. Additionally, the active trolley can introduce repeatable disturbances in its dynamic behavior by applying torque to the wheels, allowing simulation of real-world effects such as friction, irregular floors or wheel obstructions. Unaltered functionality lets the BEAST trolley be used for real actions such as loading/unloading.
 
The BEAST testbed is modular and well-documented. It comprises sensors, actuators, processing, communication and power resources, plus a ROS-based software system. Modularity allows experimenters to easily implement new benchmarks or repurpose its elements: as an example of this flexibility, the BEAST package includes an ​active wheelchair​ variant of the trolley, relevant in an ageing society scenario.
 
BEAST’s ​benchmark management software is designed to be easily integrated into external benchmarking infrastructures. It includes tools to define (via scripting, i.e., without need to modify the software) and run benchmarks based on the testbed, to apply user-defined performance metrics, and to automatically generate reports. The BEAST package comprises ​two complete benchmarks​ and a ​Gazebo testbed simulator​.
 
The BEAST consortium comprises three very experienced partners in robot benchmarking, which are also among the organisers of the ​European Robotics League1 . To promote standardisation and cross-pollination, ​the BEAST testbed and benchmarks have been proposed for integration into ERL along with the parallel MADROB testbed and benchmark, object of another FSTP-1 proposal by the same consortium. Differently from MADROB, BEAST has not been accepted for inclusion into the ERL; yet, BEAST will be able to benefit from the inclusion of MADROB in the ERL Smart CIties RObotic Challenge2 (SciRoc) taking place in September 16-22 2019. The Project consortium, in fact, will bring the prototype of the BEAST active shopping trolley and/or active wheelchair to Milton Keynes and will exploit the occasion -and the availability of ERL teams and their robots, as well as ERL organisers- for experimentation and collection of feedback. Thus, even if BEAST is not currently part of the ERL, we expect the SciRoc Challenge to provide a valuable occasion to ​validate the products of BEAST and possibly
collect datasets​. Section 4.1 is dedicated to the interactions between BEAST and the European Robotics League.