Resource-Adaptive Intelligence for Human and Robotic Systems
I lead the AIR Lab (Resource-Adaptive Intelligence for Human and Robotic Systems). The group works on
machine learning and artificial intelligence that remain useful after deployment:
adaptation under shift, reasoning and test-time computing, efficient inference on devices, and applications in mobile sensing, GenAI for Health,
robotics, and embodied AI. Our team is also actively working on Agentic AI, and more papers are coming soon.
Methodologically, the emphasis is on algorithms and systems together—what can be computed locally, under strict
resources, when data and environments change, with outputs targeted at machine learning/AI venues and top systems venues.
Resource-Adaptive & robust AI systems
Reasoning
Human-centered & embodied AI
Themes
Three focused areas around machine learning & AI at the edge, including robotics and embodied systems, with an emphasis on ML/AI-oriented conferences and journals.
Full lists: Publications and
Google Scholar.
Resource-Adaptive & robust AI systems
ML systems that stay reliable after deployment: adaptation under shift, continual improvement,
uncertainty-aware behavior, and robustness under changing users, environments, and data.
Resource-Adaptive AI · robustness · uncertainty · continual learning
Reasoning and Test Time Computing
Methods for improving reasoning quality at inference time: ambiguity-aware reasoning,
test-time scaling/adaptation, and compute strategies that improve reliability under shift.
Test-time computing · inference-time adaptation · efficient reasoning
Human-centred & embodied AI
AI/ML methods for sensing, prediction, and adaptation in GenAI for Health, affect, and activity;
learning-based control and perception in physical-world systems (e.g., wearables, vehicles, humanoids).
GenAI for Health · affective computing
robotics
ML/AI Publications by theme
Resource-Adaptive & robust AI systems
Reasoning and Test Time Computing
Human-centred & embodied AI
Resource-Adaptive & robust AI systems
Reliable AI under shift, noise, and limited supervision:
-
TinyTTA: Efficient Test-time Adaptation via Early-exit Ensembles on Edge Devices
NeurIPS 2024
Introduces efficient test-time adaptation tailored to resource-constrained edge deployment.
-
E-BATS: Efficient Backpropagation-Free Test-Time Adaptation for Speech Foundation Models
NeurIPS 2025
Extends efficient adaptation to foundation models while removing the dependence on backpropagation at test time.
-
LifeLearner: Hardware-Aware Meta Continual Learning System for Embedded Computing Platforms
ACM SenSys 2023
Studies continual adaptation on embedded platforms with hardware-aware optimization.
-
A Framework for On-Device Uncertainty-Aware Event Detection on Microcontrollers
Pervasive and Mobile Computing 2026
Develops a practical on-device framework for reliable event detection under severe microcontroller constraints.
-
UR2M: Uncertainty and Resource-Aware Event Detection on Microcontrollers
IEEE PerCom 2024
Combines uncertainty modeling with resource awareness for efficient event detection on embedded devices.
-
Efficient and Personalized Mobile Health Event Prediction via Small Language Models
ACM MobiCom 2024
Shows how compact language models can support personalized mobile health prediction efficiently.
-
PROS: an efficient pattern-driven compressive sensing framework for low-power biopotential-based wearables with on-chip intelligence
ACM MobiCom 2022
Advances low-power wearable intelligence through compressive sensing and on-chip processing.
Reasoning and Test Time Computing
Inference-time adaptation and efficient reasoning under real deployment constraints:
-
Decoding Ambiguous Emotions with Test-Time Scaling in Audio-Language Models
ArXiv Preprint 2026
Studies how additional test-time compute improves prediction for ambiguous emotion understanding in audio-language models.
-
Disentangling Reasoning in Large Audio-Language Models for Ambiguous Emotion Prediction
ArXiv Preprint 2026
Analyzes and separates reasoning behavior in multimodal models for robust prediction under ambiguous inputs.
-
Test Time Scaling for Auditory Cognition in Audio Language Models
ICASSP 2026
Demonstrates test-time scaling strategies for stronger auditory cognition in audio language models.
-
Test Time Adaptation for Speech Emotion Recognition
ICASSP 2026
Applies test-time adaptation to improve speech emotion recognition under distribution shift.
-
AdaNODEs: Test Time Adaptation for Time Series Forecasting Using Neural ODEs
ICASSP 2026
Extends test-time adaptation to time-series forecasting using neural ODE dynamics.
Human-centred & embodied AI applications
Sensing, GenAI for Health, affect, activity, and physical-world learning:
-
HoRD: Robust Humanoid Control via History-Conditioned Reinforcement Learning and Online Distillation
ArXiv 2026
Develops robust control strategies for humanoid systems under real deployment challenges.
-
CoCoEmo: Composable and Controllable Human-Like Emotional TTS via Activation Steering
ICML 2026
Explores controllable emotional speech generation for more human-like affective interaction.
-
XTransfer: Modality-Agnostic Few-Shot Model Transfer for Human Sensing at the Edge
ICML 2026
Enables few-shot cross-modality transfer for edge-deployed human sensing models.
-
LiteFat: Lightweight Spatio-Temporal Graph Learning for Real-Time Driver Fatigue Detection
IROS 2025
Targets real-time human-state understanding in safety-critical scenarios.
-
NarrativeSense: Predicting Affective States in University Students through Smartphone Sensing and Contextual Narratives
ACM Transactions on Computing for Healthcare 2026
Combines sensing and contextual narratives to understand human affect in everyday life.