[Paper Review] MoSen: Activity Modelling in Multiple-Occupancy Smart Homes

Yuting Zhan, Hamed Haddadi|arXiv (Cornell University)|Jan 1, 2021

Context-Aware Activity Recognition Systems79 references4 citations

TL;DR

MoSen is a simulation framework that models sensor-activity interactions in multi-occupancy smart homes to optimize sensor deployment for activity recognition. By emulating synthetic resident behaviors and RTLS-based identification, it quantifies trade-offs between sensor cost, localization resolution, and labeling accuracy, enabling a cost-effective sensor selection strategy that improves system performance without increasing deployment expenses.

ABSTRACT

Smart home solutions increasingly rely on a variety of sensors for behavioral analytics and activity recognition to provide context-aware applications and personalized care. Optimizing the sensor network is one of the most important approaches to ensure classification accuracy and the system's efficiency. However, the trade-off between the cost and performance is often a challenge in real deployments, particularly for multiple-occupancy smart homes or care homes. In this paper, using real indoor activity and mobility traces, floor plans, and synthetic multi-occupancy behavior models, we evaluate several multi-occupancy household scenarios with 2-5 residents. We explore and quantify the trade-offs between the cost of sensor deployments and expected labeling accuracy in different scenarios. Our evaluation across different scenarios show that the performance of the desired context-aware task is affected by different localization resolutions, the number of residents, the number of sensors, and varying sensor deployments. To aid in accelerating the adoption of practical sensor-based activity recognition technology, we design MoSen, a framework to simulate the interaction dynamics between sensor-based environments and multiple residents. By evaluating the factors that affect the performance of the desired sensor network, we provide a sensor selection strategy and design metrics for sensor layout in real environments. Using our selection strategy in a 5-person scenario case study, we demonstrate that MoSen can significantly improve overall system performance without increasing the deployment costs.

Motivation & Objective

• To address the challenge of designing cost-effective, accurate sensor networks for multi-occupancy smart homes.
• To evaluate the impact of sensor density, localization resolution, and resident count on labeling accuracy in sensor-based activity recognition.
• To develop a practical sensor selection strategy that balances system performance and deployment cost.
• To provide a simulation framework that enables early-stage design and evaluation of sensor layouts before real deployment.
• To overcome data scarcity in multi-occupancy scenarios by generating synthetic behavior models based on real human activity traces.

Proposed method

• MoSen generates synthetic multi-occupancy behavior models using real indoor activity and mobility traces from single residents.
• It simulates sensor interactions with residents using a virtual environment that incorporates floor plans and sensor layouts.
• The framework integrates RTLS-based localization to automatically annotate sensor events with resident identities, enabling accurate identification labeling.
• It evaluates performance across varying numbers of residents (2–5), sensor densities, and localization resolutions.
• The system uses trace-driven simulations to analyze individual sensor sensitivity and contribution to overall labeling accuracy.
• A sensor selection strategy is derived based on performance-cost trade-offs, prioritizing sensors with high impact and low redundancy.

Experimental results

Research questions

• RQ1 How does the number of residents affect the labeling accuracy of sensor-based activity recognition systems?
• RQ2 What is the impact of localization resolution on identification accuracy and system cost?
• RQ3 How does sensor density influence the performance and cost-efficiency of sensor networks in multi-occupancy homes?
• RQ4 Which sensor configurations maximize labeling accuracy while minimizing deployment cost?
• RQ5 How can synthetic behavior models improve the design of real-world sensor-based activity recognition systems?

Key findings

• In a 5-person scenario, MoSen’s sensor selection strategy significantly improves overall system performance without increasing deployment costs.
• Higher localization resolution improves identification accuracy but increases cost, highlighting a key trade-off in system design.
• Sensor sensitivity varies with distance; sensors with larger detection radii are less sensitive to positioning and cover broader areas.
• The framework identifies specific sensors that contribute most to labeling accuracy, enabling targeted deployment and cost reduction.
• The performance of activity recognition is significantly affected by the number of residents, sensor layout, and localization resolution.
• MoSen enables effective simulation and design of sensor networks prior to real deployment, reducing the risk of suboptimal configurations.