A5055377546- Autonomous Vehicle Technology and Safety
- Robotics and Sensor-Based Localization
- Vehicular Ad Hoc Networks (VANETs)
- Video Surveillance and Tracking Methods
- Indoor and Outdoor Localization Technologies
- Advanced Neural Network Applications
- Privacy-Preserving Technologies in Data
- Traffic Prediction and Management Techniques
- Advanced Image and Video Retrieval Techniques
- Manufacturing Process and Optimization
- Energy Efficiency and Management
- Image and Video Quality Assessment
- Age of Information Optimization
- Advanced MIMO Systems Optimization
- Sparse and Compressive Sensing Techniques
- Building Energy and Comfort Optimization
- Distributed Sensor Networks and Detection Algorithms
- Mobile Crowdsensing and Crowdsourcing
- Opportunistic and Delay-Tolerant Networks
- IoT and Edge/Fog Computing
- Energy Efficient Wireless Sensor Networks
- Augmented Reality Applications
- 3D Surveying and Cultural Heritage
- Modular Robots and Swarm Intelligence
- Machine Learning and Data Classification
University of California, Riverside
2024-2025
Southern California University for Professional Studies
2014-2022
University of Southern California
2014-2022
Chongqing University
2015-2017
Shanghai Jiao Tong University
2012
Federated learning (FL) is a rapidly growing research field in machine learning. However, existing FL libraries cannot adequately support diverse algorithmic development; inconsistent dataset and model usage make fair algorithm comparison challenging. In this work, we introduce FedML, an open library benchmark to facilitate development performance comparison. FedML supports three computing paradigms: on-device training for edge devices, distributed computing, single-machine simulation. also...
Autonomous vehicle prototypes today come with line-of-sight depth perception sensors like 3D cameras. These are used for improving vehicular safety in autonomous driving, but have fundamentally limited visibility due to occlusions, sensing range, and extreme weather lighting conditions. To improve performance, not just vehicles other Advanced Driving Assistance Systems (ADAS), we explore a capability called Augmented Vehicular Reality (AVR). AVR broadens the vehicle's visual horizon by...
Autonomous vehicles use 3D sensors for perception. Cooperative perception enables to share sensor readings with each other improve safety. Prior work in cooperative scales poorly even infrastructure support. AutoCast scalable infrastructure-less using direct vehicle-to-vehicle communication. It carefully determines which objects based on positional relationships between traffic participants, and the time evolution of their trajectories. coordinates optimally schedules transmissions a...
Like today's autonomous vehicle prototypes, vehicles in the future will have rich sensors to map and identify objects environment. For example, many prototypes today come with line-of-sight depth perception like 3D cameras. These cameras are used for improving vehicular safety driving, but fundamentally limited visibility due occlusions, sensing range, extreme weather lighting conditions. To improve performance, not just other Advanced Driving Assistance Systems (ADAS), we explore a...
In the future, video-enabled camera will be most pervasive type of sensor in Internet Things. Such cameras enable continuous surveillance through heterogeneous networks consisting fixed systems as well on mobile devices. The challenge these is to efficient video analytics: ability process videos cheaply and quickly searching for specific events or sequences events. this paper, we discuss design implementation Kestrel, a analytics system that tracks path vehicles across network. feeds are...
Precise positioning of an automobile to within lane-level precision can enable better navigation and context-awareness. However, GPS by itself cannot provide such in obstructed urban environments. In this paper, we present a system called CARLOC for automobiles. uses three key ideas concert improve accuracy: it digital maps match the vehicle known road segments; vehicular sensors obtain odometry bearing information; crowd-sourced location estimates roadway landmarks that be detected...
In-vehicle context sensing can detect many aspects of driver behavior and the environment, such as drivers changing lanes, potholes, road grade, stop signs, these features be used to improve safety comfort, engine efficiency. In general, detecting use either onboard sensors on vehicle (car sensors) or built into mobile devices (phone carried by one more occupants, both. Furthermore, traces sensor readings from different cars, when crowd-sourced, provide increased spatial coverage well...
Automotive apps can improve efficiency, safety, comfort, and longevity of vehicular use. These achieve their goals by continuously monitoring sensors in a vehicle, combining them with information from cloud databases order to detect events that are used trigger actions (e.g., alerting driver, turning on fog lights, screening calls). However, modern vehicles have several hundred describe the low level dynamics subsystems, these be combined complex ways together information. Moreover, sensor...
Collaborative Vehicular Perception (CVP) enables connected and autonomous vehicles (CAVs) to cooperatively extend their views through wirelessly sharing sensor data. Existing CVP systems employ either a vehicle-to-vehicle (V2V) or vehicle-to-infrastructure (V2I) view exchange paradigm. In this paper, we advocate hybrid design: our developed system, Harbor, employs V2I as its fundamental underlying framework, opportunistically V2V boost the performance. (helpers) may serve relays assist other...
The enormous success of advanced wireless devices is pushing the demand for higher data rates. industry satisfying this increasing by densely deploying large numbers access points (APs). Unfortunately, unicast rates, especially in crowded scenarios, remain very low due to severe interference and time-sharing. However, one may take advantage broadcasting nature transmissions offer high multicast Motivated this, we present coordinated (Co-BCast), a system which coordinates multiple APs provide...
The machining systems that mainly consist of machine tools are numerous and used in a wide range industries. total amount energy consumption by the world is extremely high. loading loss one most important complicated parts tool processes. key acquiring acquisition coefficient, which indispensable for tools’ efficiency on-line monitoring, prediction quota customization. Up to now, coefficient obtained experimental method needs conduct large experiments comprehensive measurement obtain input...
Autonomous vehicle prototypes today come with line-of-sight depth perception sensors like 3D cameras. These are used for improving vehicular safety in autonomous driving, but have fundamentally limited visibility due to occlusions, sensing range, and extreme weather lighting conditions. To improve performance, we explore a capability called Augmented Vehicular Reality (AVR). AVR broadens the vehicle's visual horizon by enabling it wirelessly share information other nearby vehicles. We show...
Disaster and emergency response operations require rapid situational assessment of the affected area for timely efficient rescue operations. A 3D map, collected after a disaster, can provide such awareness, but constructing this map quickly is significant challenge. In paper, we explore design capability called QuickSketch that rapidly builds representations an unknown environment using crowdsourcing. employs multiple vehicles equipped with sensors (stereo cameras) to different areas...
Benefiting from expanding cloud infrastructure, deep neural networks (DNNs) today have increasingly high performance when trained in the cloud. Researchers spend months of effort competing for an extra few percentage points model accuracy. However, these models are actually deployed on edge devices practice, very often, can abruptly drop over 10% without obvious reasons. The key challenge is that there not much visibility into ML inference execution devices, and little awareness potential...
The democratization of machine learning (ML) has led to ML-based vision systems for autonomous driving, traffic monitoring, and video surveillance. However, true cannot be achieved without greatly simplifying the process collecting groundtruth training testing these systems. This collection is necessary ensure good performance under varying conditions. In this paper, we present design evaluation Satyam, a first-of-its-kind system that enables layperson launch tasks with minimal effort....
In this paper, we study the Cooperative Sensing Scheduling (CSS) problem for Cognitive Radio Network (CRN), from perspective of balance between sensing performance and energy consumption. We place in a practical scenario where both primary users (PUs) secondary (SUs) are heterogeneous: PU channels different terms channel admission control, idle probability capacity; SUs differs formulate CSS as programming problem, whose optimal solution is proved to exist but takes considerable time reach....
To develop medical expert systems (MES), researchers and practitioners usually apply a machine learning (ML) classifier they expect to be the best through ML effect estimation. A standard can perform well on balanced datasets, but its performance often dramatically decreases when applied unbalanced datasets. address these challenges, this paper proposes hybrid sampling approach for evaluating classifiers with data in MES. In framework, mean responses of classification experiments are...
Efficient point cloud (PC) compression is crucial for streaming applications, such as augmented reality and cooperative perception. Classic PC techniques encode all the points in a frame. Tailoring towards perception tasks at receiver side, we ask question, "Can remove ground during transmission without sacrificing detection performance?" Our study reveals strong dependency on from state-of-the-art (SOTA) 3D object models, especially those below around object. In this work, propose...