Leifeng.com AI Technology Review: ICRA stands for "IEEE International Conference on Robotics and Automation", which is one of the most influential international academic conferences in the field of robotics. ICRA 2017 was held from May 29 to June 4, and Leifeng.com AI Technology Review brought first-hand reports from Singapore. During the conference, Leifeng.com will launch a series of special reports around the conference agenda and award-winning papers, so stay tuned.

Below is the abstract of the paper that won the ICRA 2017 Best Service Robotics Paper.

High-precision microinjection of microbeads into Caenorhabditis elegans trapped in a suction microchannel

This study demonstrated high-precision microinjection of fluorescent microgel beads into Caenorhabditis elegans trapped in an aspiration microchannel. In previous work by the team, they demonstrated survival microinjection by conventional micromanipulation techniques. However, the focal planes between the tip of the microinjection tool and the target axon in the nematode body were different. To address this problem, they proposed an aspiration microchannel that captures the nematode during microinjection manipulation. The focal plane of the target axon matched that of the fluorescent microbeads in the microinjection tool, enabling high-precision microinjection into the nematode body under microscopic view. In experimental evaluation, the localization accuracy of injection into the nematode was within the target accuracy (15m). The first navigational positioning of the nematode along the microchannel was controlled by electroosmosis. The injection of fluorescent microgel beads into the nematode body was quantitatively confirmed by confocal microscopy.

Cornell University team proposes an evaluation method for autonomous robotic systems to perform non-destructive bridge deck inspections

Bridge condition assessment is important to maintain the quality of public highways. Deterioration of bridges due to time is inevitable due to aging of materials, environmental wear and tear, and in some cases poor maintenance. Nondestructive evaluation (NDE) methods are a preferred approach for condition assessment of bridges, concrete buildings, and other civil structures. Some examples of NDE methods are ground penetrating radar (GPR), acoustic emission, and electrical resistivity (ER). NDE methods provide the ability to inspect structures without causing any damage to the structure during the inspection process. In addition, the cost of NDE methods is generally lower than other methods because they do not require evacuation of the inspection point prior to inspection, which significantly reduces the cost of safety-related issues during the inspection process. This paper proposes an autonomous robotic system equipped with three different NDE sensors. The system employs GPR, ER, and cameras for data acquisition. The system is capable of real-time, cost-effective bridge deck inspection and includes mechanical robotic design and machine learning and pattern recognition methods for automatic rebar selection to provide a real-time condition map of the corrosive deck environment.

Kyushu University team proposes feasibility study of "Big Sensor Box" platform

This paper proposes a new software and hardware platform for an information-based structured environment named ROS-TMS and Big Sensor Box. The team has developed an information-based structured environment management system named TMS (Town Management System) since 2005 in the Robot City project. Since then, they have continued to work on improving the functionality and performance of TMS. Recently, they launched the latest version of TMS named ROS-TMS, which solves some key issues in TMS by adopting ROS (Robot Operating System) and taking advantage of high scalability and a large number of ROS resources. In this paper, the structure of the software platform for the information-based structured environment will first be discussed, and the latest system ROS-TMS version 4.0 will be described in detail. Next, they introduce a hardware platform for an information-based structured environment named Big Sensor Box, in which various sensors are embedded and the service robot is operated according to the structured information managed by ROS-TMS. Robot service experiments including acquisition tasks and autonomous control of wheelchair robots are also carried out in the Big Sensor Box.

The Hong Kong University of Science and Technology team proposed using environmental sparsity to improve occtree-based occupancy maps for aerial robot navigation applications

In this paper, the team proposed an improved octree-based mapping framework for autonomous navigation of mobile robots. Octrees are known for their memory efficiency in representing large-scale environments. However, existing implementations including state-of-the-art octree maps are computationally too expensive for online applications that require frequent map updates and queries. Exploiting the sparsity of the environment, the team proposed a ray tracing method with early termination for efficient probabilistic map updates. They also proposed a partition and volume occupancy query method as the core operation to generate free space configurations based on optimization-based trajectory generation. Their experiments demonstrated that the method maintains the same storage advantage as the original octree map, but is computationally more efficient for map updates and occupancy queries. Finally, an autonomous quadrotor flight through a complex environment is demonstrated by integrating the proposed map structure into a complete navigation pipeline.

Original link: A review of the best service robot papers from ICRA 2017