Rescue robot Quince is specialized to facilitate a smooth disaster response activity and minimize the human suffering from the secondary disaster by investigating a hazardous disaster site with its cutting edge locomotion and sensing functions instead of human responders. Quince can negotiate staircases and bumpy ground by utilizing track belts covering its main body and four sub-tracks swinging like arms, and be optionally equipped with sensors and a manipulator for a specific mission. As add-on functions for Quince, we have developed tele-operation assistance for a unskilled operator and 3D map building based on sensing the environment.

We have developed an autonomy of Quince with sub-tracks which facilitates a tele-operation. In practical use, the operator conducts Quince from far distance when Quince explores a disaster site. However it is difficult to maneuver Quince safely in partially-collapsed building which contains staircases and bumps with the traditional tele-operation. Thus we have developed an autonomy which compensates a part of tele-operation. It decreases the operator’s workload for maneuvering Quince. With the autonomy, the operator can maneuver Quince through staircases and bumps just by specifying the moving direction. This tele-operation assist system was verified in actual facilities such as underground mall (Sannomiya, Hyogo, Japan) and training facilities of simulated rubbles.

We developed distributed tactile sensors for tracked vehicle that can measure the positions and the pressures of contacts to the ground with respect to the track surface. Contact information is significantly important for the tracked vehicles which move on rubbles. The tracked vehicle can avoid rollover and get over rubbles on the basis of contact information. The problems of distributing tactile sensors among the track surface are a sensing method, configuration, and wiring. As solutions for those issues, we are developing multiple types of distributed tactile sensors for tracking.

Wide range scan in real time (3D map_movie)

Scanning detailed object shape （3D map_movie）

We have researched on a 3D measurement by a laser scanner attached on a mobile robot. By collecting detailed information of a large environment, we can utilize the information to improve the ease of the tele-operation, and clearly show the condition of the environment explored by the robot to operators.

Categorization of objects in 3D Point Cloud (movie)

To clearly show what is included in a 3D point cloud obtained from a laser scanner, we have researched on the extraction and the categorization of objects from a 3D point cloud such as polls, grounds, walls, rubbles, and etc. This categorization is based on the shape of point cloud in each voxel and the passing rate against the laser. By colorizing objects according to the categorization result, operators understand the environment around the robot intuitively.

We developed a simulator of a tracked vehicle with sub-tracks named Kenaf, which is aimed to train the tele-operation skill of an operator, and to help to develop sensing methods and software components of autonomous control for the tracked vehicle. It simulates the locomotion of the tracked vehicle in various 3D environments so accurately that developers can verify how their software works with the tracked vehicle without setting up the physical one. It is based on the mobile robot simulator USARSim that is used worldwide.

We develop hardware/software of a multi-camera system for rescue robots, which is inspired by a compound eye of an insect. The system contains multi CMOS cameras and a FPGA which fast processes images from the cameras. The user can watch every processed image from the system via Ethernet. The final goal of this research is to create a robot equipped with dozens of distributed cameras.

We have aimed to develop a multi-modal interface for tele-operation. Operators controlled mobile robots using visual and auditory information during tele-operation. However, it was hard for the operator to sense force information which is important to control the mobile robots on rubble. Therefore, we developed a joystick with force display. We displayed virtual posture information to the operators by using the joystick.

As a disaster response for the Fukushima Daiichi nuclear power plant disaster resulted from the Tohoku earthquake on 11 March 2011, several specialized versions of rescue robot Quince, was developed by Tohoku University including our laboratory, Chiba Institute of Technology, and International Rescue System (IRS), were utilized to explore a multi-story reactor building seriously damaged.