Object detection is a widely-used computer vision task, in which we identify a bounding box around each object in an image and classify the object into one of pre-defined classes. Single Shot MultiBox Detector (SSD) is a real-time object detector based on a single convolutional neural network. SSD is popular and known for high speed and accuracy, but its black-box nature is not ignorable when it is applied to critical systems. In this paper, we propose Contrastive Relevance Propagation (CRP), an extension of Layer-wise Relevance Propagation (LRP) tailored for SSD. CRP can consistently deal with SSD's heterogeneous output, i.e. confidences for object classes and location offsets, and create a heatmap that highlights a crucial part of the input, which is not available with a standard use of LRP. By experiments with the Pascal VOC 2012 dataset, we confirmed the quality of heatmaps created by CRP, and with such heatmaps, we conducted a simple analysis on prediction errors made by SSD.

In this paper, we propose a new algorithm to control a distributed autonomous robotic system under dynamically changing environment based on immunological interaction between robots. Our algorithm can organize the robot population for dynamically changing multiple works. At first, we designed a control architecture for a multiple robotic system based on B-cell (which is main agent of immune system) interaction. Immune system has various kinds of B-cells, and B-cell interaction can organize its population balance against dynamically changing environment. We set the analogy between distributed autonomous robotic system and biological immune system. We verified the performance of our algorithm by computer simulation. As a simulation example there is considered the transportation of multiple objects to multiple locations against a deadline, with time-varying demand.

This paper proposes a population control architecture for a multiple robotic system inspired from biological immune networks. The authors' architecture organizes suitable population balance against dynamic environment independently of any centralized control, while each robot decides its action without a global world model. This group behavior, suitable population balance, is organized as side-effect of the interaction of the individual robots in the world.

This paper presents both the hardware and the software architectures for the multi agent robotic system. For the hardware architecture of the multi agent robotic system, the authors show the programmable MARS (micro autonomous robotic system). This robot can work for one of the agents of the multi agent robotic system. The size of this robot is 20 mm cubic. The authors can rewrite and download any programs by the infrared communication function and the programmable function. For the software architecture of the multi agent system, the authors propose the algorithm which can adapt and generate the swarm strategy in the dynamic environment, based on the biological immune system. The authors adopt the self and non-self recognition network of the immune system in their algorithm.