Dadu-CD: Fast and Efficient Processing-in-Memory Accelerator for Collision Detection

Yang, Yuxin; Chen, Xiaoming; Han, Yinhe

doi:10.1109/dac18072.2020.9218709

Cited by 12 publications

(11 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Thus, approximate methods such as probabilistic roadmaps [55] are widely used [61], [90] over the last 20 years. Many hardware accelerators [6], [66], [76], [77], [93], [116] have been proposed for these algorithms.…”

Section: Motion Planningmentioning

confidence: 99%

“…Leven and Hutchinson [62] proposed a real-time motion planning approach based on probabilistic roadmaps for a dynamic environment and is also used by MPAs [39], [66], [76], [77], [93], [116]. In this approach, the motion set is generated for an obstaclefree environment.…”

Section: Probabilistic Roadmapsmentioning

confidence: 99%

“…Unfortunately, an erroneous collision detection can potentially lead to a collision between the robot and an obstacle in its surroundings, making collision detection safety-critical. Thus, we focus on the CDM and perform fault characterization of CDMs of four MPAs, which we refer to as A1 [78], A2 [76], A3 [66], and A4 [116] (Section 5.2) throughout this paper. Collision detection process identifies this motion to be in collision.…”

Section: Probabilistic Roadmapsmentioning

confidence: 99%

“…In contrast, GPU-based collision detection acceleration [10], [29] provides high flexibility, but it is not energyefficient. Configurable collision detection hardware accelerators [6], [66], [76], [77], [93], [116] provide a balance between flexibility and performance/energy-efficiency. We focus on such configurable CDM architectures.…”

Section: Collision Detection Modules (Cdm)mentioning

confidence: 99%

“…The computational and real-time demands of motion planning exceed those provided by typical CPUs. Recently, several approaches have been proposed to accelerate motion planning on different hardware platforms, including GPUs [10], [29], FPGAs [6], [78], [93], and ASICs [66], [76], [95], [116]. Motion Planning Accelerators (MPAs) have achieved impressive performance gains and are being adopted in industry [46], [84], [89].…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Characterizing and Improving the Resilience of Accelerators in Autonomous Robots

Shah¹,

Xue²,

Pattabiraman³

et al. 2021

Preprint

View full text Add to dashboard Cite

Motion planning is a computationally intensive and well-studied problem in autonomous robots. However, motion planning hardware accelerators (MPA) must be soft-error resilient for deployment in safety-critical applications, and blanket application of traditional mitigation techniques is ill-suited due to cost, power, and performance overheads. We propose Collision Exposure Factor (CEF), a novel metric to assess the failure vulnerability of circuits processing spatial relationships, including motion planning. CEF is based on the insight that the safety violation probability increases with the surface area of the physical space exposed by a bit-flip. We evaluate CEF on four MPAs. We demonstrate empirically that CEF is correlated with safety violation probability, and that CEF-aware selective error mitigation provides 12.3×, 9.6×, and 4.2× lower Failures-In-Time (FIT) rate on average for the same amount of protected memory compared to uniform, bit-position, and access-frequency-aware selection of critical data. Furthermore, we show how to employ CEF to enable fault characterization using 23, 000× fewer fault injection (FI) experiments than exhaustive FI, and evaluate our FI approach on different robots and MPAs. We demonstrate that CEF-aware FI can provide insights on vulnerable bits in an MPA while taking the same amount of time as uniform statistical FI. Finally, we use the CEF to formulate guidelines for designing soft-error resilient MPAs.

show abstract