1.  Application Development for Real-World Intelligent Systems
  A real-world intelligent system consists of three basic modules as shown in Fig.1: environment recognition, prediction or estimation, and behavior planning. We study high-quality processing algorithms for the real-world applications such as a highly-safe intelligent system, a medical information support system and a big-data processing system as shown in Fig.1. Moreover, we are developing a platform that from a set of training samples learns to select the best algorithm, so that it provides automatic and adaptive algorithm selection capability according to real-world environment conditions as shown in Fig.2(a).

  2.  High-performance Low-Power Computing Platform
  To achieve very high-performance and low-power computation, high-level synthesis including optimization techniques such as minimization of a number of data transfers between memories and functional units is investigated. Moreover, a heterogeneous computing platform with CPUs, GPUs and FPGA custom accelerators are developed for high-performance and low-power computing. Automatic mapping methodology for a user program is developed toward the super computing.

  3.  New-Concept Reconfigurable VLSI Computing
  Reconfigurable VLSIs superior to the conventional FPGAs are designed and implemented based on new-concept architectures such as power gating with asynchronous control, an asynchronous/synchronous hybrid architecture, a packet data transfer scheme, and a multiple-valued logic-in-memory architecture as shown in Fig.2(b).