arXiv subjects

Eli Bozorgzadeh

Publications and source records attributed to Eli Bozorgzadeh.

6 recordsLinked to original sources

arXiv2026

FRESHLATENT: Channel-Aware Latent Adaptation for Resource-Constrained Embodied VLM Perception

Mission-critical UAVs increasingly rely on split vision-language model (VLM) perception under tight onboard-resource and wireless-communication constraints. However, corruption of transmitted intermediate features creates a deployment mismatch for clean-trained split interfaces, while stronger channel-aware codecs can impose substantial onboard cost. We present FreshLatent, a lightweight channel-aware latent adapter that trains a power-normalized encoder-decoder through wireless corruption while keeping the surrounding VLM frozen. We formulate deployment around a mission-conditioned perception requirement and embedded interface cost, linking channel quality and communication budget to the operating conditions under which perception remains usable. At 0 dB and the tightest communication budget, FreshLatent improves gIoU and cIoU over clean split compression by 20.79 and 20.87 points, respectively. At the most adverse evaluated SNR (0 dB), across all three communication budgets, FreshLatent recovers 63.5-69.1% of the gIoU improvement achieved by a much heavier, range-trained feature-JSCC codec. On an NVIDIA Jetson AGX Xavier in 10-W mode, FreshLatent uses 37-40x fewer encoder parameters, 7.7-9.9x lower edge-interface latency, and 8.8-10.0x lower edge-interface energy than the heavier codec. Together, these results show that lightweight channel-aware adaptation can recover a substantial fraction of the robustness of a much larger communication interface while broadening quality-valid operation under constrained wireless conditions.

eess.SP↗

arXiv2026

FloodReasonBench: Benchmarking VLM Reasoning Segmentation for Embodied Flood Response at the Edge

Reasoning segmentation enables vision-language models (VLMs) to translate mission-relevant language requests into pixel-level visual grounding, offering a natural perception interface for embodied agents. However, existing benchmarks largely focus on generic visual scenes and overlook the domain and resource constraints encountered in flood-response platforms. We present FloodReasonBench, a benchmark for VLM reasoning segmentation for embodied flood response at the edge. At its core, FloodReasonBench introduces FloodResponseSeg, a flood-specific reasoning-segmentation dataset constructed from real-world scenes and response-relevant targets. Beyond task accuracy, the benchmark characterizes reasoning-segmentation pipelines under lightweight visual encoding, hierarchical split inference, and compressed intermediate representations. We observe strong partition-dependent accuracy variation in the generic pre-adaptation setting, while the flood-adapted target-workload design space exhibits a substantially more compact accuracy range across partitions. Evaluation on an NVIDIA Jetson AGX Xavier further exposes the tradeoffs among reasoning-segmentation accuracy, edge-side latency, energy, and communication footprint, enabling quality-constrained selection of edge operating points. Together, these results provide a task- and system-level characterization of reasoning segmentation for resource-constrained embodied flood response at the edge.

cs.DC↗

arXiv2026

MemoGuard: An Adaptive Runtime for Guarding Against Memory Traps in Communication-Limited Robot Navigation

Communication-limited robots in mission-critical scenarios such as disaster inspection and search-and-rescue must make reliable onboard decisions without access to remote operators or high-capacity reasoning services. Episodic memory reuse is an attractive low-cost fallback, but retrieval similarity does not guarantee execution validity, i.e., a retrieved action may match the current context yet be unsafe due to changed topology, insufficient battery margin, or unreliable prior outcomes. We call such high-similarity but execution-invalid episodes memory traps. This creates a safety-efficiency design space where similarity only reuse minimizes fallback cost but can be unsafe, while always invoking local reasoning improves safety at high computational and energy cost. This paper presents MemoGuard, a lightweight adaptive runtime that validates episodic memories against topology, resource, and outcome contracts before reuse, invoking fallback only when validation fails. In a graph-based corridor-inspection simulator, MemoGuard reduces battery safety violations by 76.6% over similarity-only top-1 reuse while reducing fallback calls by 21.4% over always reasoning. On an NVIDIA Jetson AGX Xavier with local llama3.2:3b fallback reasoning, this corresponds to 3.67 s and 36.97 J of avoided fallback-reasoning overhead per trial. We open-source MemoGuard at https://github.com/hetheiin/memoguard.

cs.RO↗

arXiv2026

Terastal: Layer-Variant-based Scheduling for Real-Time Multi-DNN Workloads on Heterogeneous Accelerators

Heterogeneous DNN accelerators improve soft real-time multi-DNN execution by mapping each layer to its preferred accelerator to reduce latency. However, under skewed workloads, large layer-latency differences across accelerators limit scheduling flexibility and increase deadline misses. To address this challenge, we introduce layer variants, customized layer implementations that reduce latency gaps on non-preferred accelerators. We then present Terastal, a soft real-time framework for layer-variant design and scheduling on heterogeneous DNN accelerators. Terastal combines offline heterogeneity-aware virtual budget assignment and layer-variant design, and online scheduling to jointly optimize accelerator mapping and variant selection under timing and accuracy constraints. Experimental results show that Terastal reduces deadline miss rate per model by 40.58%, 30.53%, and 36.27% compared with FCFS, EDF, and DREAM, respectively, while incurring only 2.24% average normalized accuracy loss across models with variants.

cs.DC↗

arXiv2019

UltraShare: FPGA-based Dynamic Accelerator Sharing and Allocation

Despite all the available commercial and open-source frameworks to ease deploying FPGAs in accelerating applications, the current schemes fail to support sharing multiple accelerators among various applications. There are three main features that an accelerator sharing scheme requires to support: exploiting dynamic parallelism of multiple accelerators for a single application, sharing accelerators among multiple applications, and providing a non-blocking congestion-free environment for applications to invoke the accelerators. In this paper, we developed a scalable fully functional hardware controller, called UltraShare, with a supporting software stack that provides a dynamic accelerator sharing scheme through an accelerators grouping mechanism. UltraShare allows software applications to fully utilize FPGA accelerators in a non-blocking congestion-free environment. Our experimental results for a simple scenario of a combination of three streaming accelerators invocation show an improvement of up to 8x in throughput of the accelerators by removing accelerators idle times.

cs.AR↗

arXiv2015

Modeling And Control Battery Aging in Energy Harvesting Systems

Energy storage is a fundamental component for the development of sustainable and environment-aware technologies. One of the critical challenges that needs to be overcome is preserving the State of Health (SoH) in energy harvesting systems, where bursty arrival of energy and load may severely degrade the battery. Tools from Markov process and Dynamic Programming theory are becoming an increasingly popular choice to control dynamics of these systems due to their ability to seamlessly incorporate heterogeneous components and support a wide range of applications. Mapping aging rate measures to fit within the boundaries of these tools is non-trivial. In this paper, a framework for modeling and controlling the aging rate of batteries based on Markov process theory is presented. Numerical results illustrate the tradeoff between battery degradation and task completion delay enabled by the proposed framework.

eess.SY↗