Ushasi Ghosh
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Ushasi Ghosh has a total of 20 co-authors. Here are the top co-authors by number of shared publications:
- Dinesh Bharadia
University of California San Diego, La Jolla, CA, USA
13 shared publications - Srinivas Shakkottai
Texas A&M University, College Station, USA
10 shared publications - Qingyuan Zheng
University of California San Diego, San Diego, CA, USA
3 shared publications - Ali Mamaghani
University of California San Diego, La Jolla, USA
2 shared publications - Archana Bura
Postdoctoral Scholar, UCSD, San Diego, USA
2 shared publications - Azuka Chiejina
Electrical and Computer Engineering, George Mason University, Fairfax, Virginia, United States
2 shared publications
Latest Publications
- Open Access
- 22 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 351-360
Windex: Index-Driven Contextual Scheduler for Scalable Service Guarantees in NextG Cellular Networks
Next-generation cellular networks must support heterogeneous service objectives, ranging from throughput-oriented eMBB to latency- and regularity-constrained URLLC and XR traffic. Enforcing these objectives at the RAN scheduler is difficult because wireless channels, queues, and application arrivals evolve at the millisecond timescale while spectrum resources are shared across users. We formulate service-aware RAN scheduling as a constrained Markov decision process with per-user throughput and service-regularity constraints. A Lagrangian relaxation gives an exact decomposition of the relaxed problem into per-user control problems with a common resource price. We prove fixed-dual indexability for an illustrative time-since-last-service (TSLS) debt arm with a throughput objective and TSLS constraint, note the analogous scalar throughput-debt result, and assume fixed-dual indexability for the full vector state used in deployment. We design Windex, an index-driven scheduler that learns dual-dependent per-user break-even resource prices using compact neural approximators over queue, channel, TSLS, and violation-deficit state. We further design CWindex, a contextual extension that infers service class from queue-history time series when explicit labels are unavailable. We implement Windex as an application compatible with a realtime 5G radio intelligent controller. Inference takes 10-20 μs per user on a general-purpose CPU, well within a 1 ms control budget. Trace-driven emulation using over-the-air channel measurements and larger simulations show that Windex and CWindex substantially reduce per-user service violations while maintaining throughput efficiency.
- 22 November 2026
- Published by Association for Computing Machinery (ACM)
- p. 351-360
Open AccessWindex: Index-Driven Contextual Scheduler for Scalable Service Guarantees in NextG Cellular Networks
Next-generation cellular networks must support heterogeneous service objectives, ranging from throughput-oriented eMBB to latency- and regularity-constrained URLLC and XR traffic. Enforcing these objectives at the RAN scheduler is difficult because wireless channels, queues, and application arrivals evolve at the millisecond timescale while spectrum resources are shared across users. We formulate service-aware RAN scheduling as a constrained Markov decision process with per-user throughput and service-regularity constraints. A Lagrangian relaxation gives an exact decomposition of the relaxed problem into per-user control problems with a common resource price. We prove fixed-dual indexability for an illustrative time-since-last-service (TSLS) debt arm with a throughput objective and TSLS constraint, note the analogous scalar throughput-debt result, and assume fixed-dual indexability for the full vector state used in deployment. We design Windex, an index-driven scheduler that learns dual-dependent per-user break-even resource prices using compact neural approximators over queue, channel, TSLS, and violation-deficit state. We further design CWindex, a contextual extension that infers service class from queue-history time series when explicit labels are unavailable. We implement Windex as an application compatible with a realtime 5G radio intelligent controller. Inference takes 10-20 μs per user on a general-purpose CPU, well within a 1 ms control budget. Trace-driven emulation using over-the-air channel measurements and larger simulations show that Windex and CWindex substantially reduce per-user service violations while maintaining throughput efficiency.
- Open Access
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1900-1902
Demo: RoboNEST: Robotics Networked Edge Systems Testbed for Physical AI
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1900-1902
Open AccessDemo: RoboNEST: Robotics Networked Edge Systems Testbed for Physical AI
- Open Access
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1677-1680
WiP: mmsRAN: Multimodal Cellular Uplink for Physical AI
Physical AI is increasingly relying on offloading the multimodal perception and planning to the cellular edge, placing the cellular uplink directly in the robot's control loop. Each control iteration generates a group of correlated sensor streams that must reach the edge together for timely inference. Existing cellular uplink mechanisms, however, treat these streams as independent packet flows, introducing grant-access latency and synchronization loss across modalities and users. We present mmsRAN, an application-aware uplink architecture that treats a robot's multimodal sensor streams as a single control frame. mmsRAN combines proactive uplink scheduling with per-modality Data Radio Bearers (DRBs) to reduce access latency while prioritizing latency-critical sensor streams under constrained uplink resources. Experiments on a private-5G robotic testbed demonstrate up to 37% lower sensor-frame latency and over 2× lower latency for high-priority sensor streams.
- 26 October 2026
- Published by Association for Computing Machinery (ACM)
- p. 1677-1680
Open AccessWiP: mmsRAN: Multimodal Cellular Uplink for Physical AI
Physical AI is increasingly relying on offloading the multimodal perception and planning to the cellular edge, placing the cellular uplink directly in the robot's control loop. Each control iteration generates a group of correlated sensor streams that must reach the edge together for timely inference. Existing cellular uplink mechanisms, however, treat these streams as independent packet flows, introducing grant-access latency and synchronization loss across modalities and users. We present mmsRAN, an application-aware uplink architecture that treats a robot's multimodal sensor streams as a single control frame. mmsRAN combines proactive uplink scheduling with per-modality Data Radio Bearers (DRBs) to reduce access latency while prioritizing latency-critical sensor streams under constrained uplink resources. Experiments on a private-5G robotic testbed demonstrate up to 37% lower sensor-frame latency and over 2× lower latency for high-priority sensor streams.
- Open Access
- 25 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 19-24
Scout: User Plane Telemetry for NextG Cellular Networks
- 25 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 19-24
Open AccessScout: User Plane Telemetry for NextG Cellular Networks
- Open Access
- 25 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 7-12
Beyond Wi-Fi: Private 5G for Better Multi-User Quality-of-Experience
- 25 February 2026
- Published by Association for Computing Machinery (ACM)
- p. 7-12
Open AccessBeyond Wi-Fi: Private 5G for Better Multi-User Quality-of-Experience
- Open Access
- 4 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 36-42
Log Pruning for Efficient Q&A in 5G Networks
- 4 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 36-42
Open AccessLog Pruning for Efficient Q&A in 5G Networks
- Open Access
- 4 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 7-9
PHOENIX: Protocol State Machine Recovery from Events in 5G Logs
- 4 November 2025
- Published by Association for Computing Machinery (ACM)
- p. 7-9
Open AccessPHOENIX: Protocol State Machine Recovery from Events in 5G Logs
- Open Access
- 4 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 2312-2314
DEMO: SPARC: Spatio-Temporal Adaptive Resource Control for Multi-site Spectrum Management in NextG Cellular Networks
This work presents SPARC (Spatio-Temporal Adaptive Resource Control), a novel approach for multi-site spectrum management in NextG cellular networks. SPARC addresses the challenge of limited licensed spectrum in dynamic environments. We leverage the O-RAN architecture to develop a multi-timescale RAN Intelligent Controller (RIC) framework, featuring an xApp for near-real-time interference detection and localization, and a μApp for real-time intelligent resource allocation. By utilizing base stations as spectrum sensors, SPARC enables efficient and fine-grained dynamic resource allocation across multiple sites, enhancing signal-to-noise ratio (SNR) by up to 7dB, spectral efficiency by up to 15%, and overall system throughput by up to 20%.
- 4 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 2312-2314
Open AccessDEMO: SPARC: Spatio-Temporal Adaptive Resource Control for Multi-site Spectrum Management in NextG Cellular Networks
This work presents SPARC (Spatio-Temporal Adaptive Resource Control), a novel approach for multi-site spectrum management in NextG cellular networks. SPARC addresses the challenge of limited licensed spectrum in dynamic environments. We leverage the O-RAN architecture to develop a multi-timescale RAN Intelligent Controller (RIC) framework, featuring an xApp for near-real-time interference detection and localization, and a μApp for real-time intelligent resource allocation. By utilizing base stations as spectrum sensors, SPARC enables efficient and fine-grained dynamic resource allocation across multiple sites, enhancing signal-to-noise ratio (SNR) by up to 7dB, spectral efficiency by up to 15%, and overall system throughput by up to 20%.
- Open Access
- 4 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1823-1825
Demo: Realtime Neural Whittle Indexing for Scalable Service Guarantees in NextG Cellular Networks
This work presents Windex, a novel light weight whittle index network-driven realtime scheduler for scalable service guarantees in NextG cellular networks. Windex addresses the resource allocation challenge in NextG cellular radio access networks (RAN), where resources must be shared among diverse user applications, each requiring guarantees on throughput and service regularity, taking into account service guarantees, channel quality, and system load. Implemented in a real time intelligent controller (RIC), and evaluating across standardized 3GPP service classes, we demonstrate the least service violations compared to state-of-the-art systems using over-the-air channel traces on a 5G testbed.
- 4 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1823-1825
Open AccessDemo: Realtime Neural Whittle Indexing for Scalable Service Guarantees in NextG Cellular Networks
This work presents Windex, a novel light weight whittle index network-driven realtime scheduler for scalable service guarantees in NextG cellular networks. Windex addresses the resource allocation challenge in NextG cellular radio access networks (RAN), where resources must be shared among diverse user applications, each requiring guarantees on throughput and service regularity, taking into account service guarantees, channel quality, and system load. Implemented in a real time intelligent controller (RIC), and evaluating across standardized 3GPP service classes, we demonstrate the least service violations compared to state-of-the-art systems using over-the-air channel traces on a 5G testbed.
- Open Access
- 4 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1587-1589
AppNet: Application-Aware Networking with O-RAN
The 5G network promised transformative services across various industries, yet its integration has mostly been limited to existing 4G services like IMS-based multimedia and IoT. This paper identifies two key reasons for this underutilization: first, the stringent, multi-dimensional requirements of next-generation verticals like AR/VR, mobile gaming, and robotics, and the limitations of traditional Quality of Service (QoS) approaches in meeting these needs. We argue for a shift towards Quality of Experience (QoE), which better captures user perception and instantaneous application state. To meet stringent QoE demands and optimize network utilization, we emphasize the importance of application state and context awareness within the network. As a solution, we propose AppNet, a novel framework that integrates application awareness into the networking stack via RAN Intelligent Controllers (RICs) of the Open-RAN platform, enabling dynamic QoS adjustments based on application context. This paper highlights 5G private networks as an ideal testing ground, focusing on multi-user scenarios to deliver optimal real-time interactive services at scale.
- 4 December 2024
- Published by Association for Computing Machinery (ACM)
- p. 1587-1589
Open AccessAppNet: Application-Aware Networking with O-RAN
The 5G network promised transformative services across various industries, yet its integration has mostly been limited to existing 4G services like IMS-based multimedia and IoT. This paper identifies two key reasons for this underutilization: first, the stringent, multi-dimensional requirements of next-generation verticals like AR/VR, mobile gaming, and robotics, and the limitations of traditional Quality of Service (QoS) approaches in meeting these needs. We argue for a shift towards Quality of Experience (QoE), which better captures user perception and instantaneous application state. To meet stringent QoE demands and optimize network utilization, we emphasize the importance of application state and context awareness within the network. As a solution, we propose AppNet, a novel framework that integrates application awareness into the networking stack via RAN Intelligent Controllers (RICs) of the Open-RAN platform, enabling dynamic QoS adjustments based on application context. This paper highlights 5G private networks as an ideal testing ground, focusing on multi-user scenarios to deliver optimal real-time interactive services at scale.