We present an accurate and easy-to-use multi-sensor fusion toolbox for autonomous vehicles. It includes a ‘target-less’ multi-LiDAR (Light Detection and Ranging), and Camera-LiDAR calibration, sensor fusion, and a fast and accurate point cloud ground classifier. Our calibration methods do not require complex setup procedures, and once the sensors are calibrated, our framework eases the fusion of multiple point clouds, and cameras. In addition we present an original real-time ground-obstacle classifier, which runs on the CPU, and is designed to be used with any type and number of LiDARs. Evaluation results on the KITTI dataset confirm that our calibration method has comparable accuracy with other state-of-the-art contenders in the benchmark.

This paper presents a parallelization method utilizing dead time to implement higher precision feedback control systems in multicore processors. The feedback control system is known to be difficult to parallelize, and it is difficult to deal with the dead time in control systems. In our method, the dead time is explicitly represented as delay elements. Then, these delay elements are distributed to the overall systems with equivalent transformation so that the system can be simulated or executed in parallel pipeline operation. In addition, we introduce a method of delay-element addition for parallelization. For a spring-mass-damper model with a dead time, parallel execution of the model using our technique achieves 3.4 times performance acceleration compared with its sequential execution on an ideal four-core simulation and 1.8 times on a cycle-accurate simulator of a four-core embedded processor as a threaded application on a real-time operating system.

This paper presents an algorithm for the scan-chain optimization problem in multiple-scan design methodology. The proposed algorithm, which consists of four phases, first determines pairs of scan-in and scan-out pins (Phase 1), and then assigns flip-flops to scan-paths by using a graph theoretical method (Phase 2). Next the algorithm decides connection-order of flip-flops in each scan-path by using TSP (Traveling Salesman Problem) heuristics (Phase 3), and finally exchanges flip-flops among scan-paths in order to reduce total scan-path length (Phase 4). Experiments using actual design data show that, for ten scan-paths, our algorithm achieved a 90% reduction in scan-test time at the expense of a 7% total scan-path length increase as compared with the length of a single optimized scan-path. Also, our algorithm produced less total scan-path length than other three possible algorithms in a reasonable computing time.

We have proposed area-reduction techniques for superscalar datapath architectures with 34 SIMD instructions and have developed an integer-media unit based on these techniques. The unit's design is both functionally asymmetrical and integer-SIMD unified, and the resulting savings in area are 27%-48% as compared to other, functionally equivalent mid-level microprocessor designs, with performance that is, at most, only 7.2% lower. Further, in 2-D IDCT processing, the unit outperforms embedded microprocessor designs without SIMD functions by 49%-118%. Specifically, effective area reduction of adders, shifters, and multiply-and-adders has been achieved by using the unified design. These area-effective techniques are useful for embedded microprocessors and scalable systems that employ highly parallel superscalar and on-chip parallel architectures. The integer-media unit has been implemented in an evaluation chip fabricated with 0.15-µm 5-metal CMOS technology.

The planar point location problem is one of the most fundamental problems in computational geometry and stated as follows： Given a straight line planar graph (subdivision) with n vertices and an arbitary query point Q, determine the region containing Q. Many algorithms have been proposed, and some of them are known to be theoretically optimal (O(log n) search time, O(n) space and O(n log n) preprocessing time). In this paper, we implement several representative algorithms in C, and investigate their practical efficiencies by computational experiments on Voronoi diagrams with 210 - 217 vertices.