In a recent study, Life Adapt CRO Diane Cook in collaboration with researchers at Washington State University offers an innovative exploration of human mobility within indoor environments. This research is significant because, unlike outdoor human mobility, which has been extensively studied, indoor human mobility has not received as much attention. Yet, it’s crucial for enhancing smart home and city services. The paper is a pioneering effort to theoretically estimate and empirically validate the predictability of human movement in residential settings using smart home sensor data.
Study Objective and Data Collection
The objective of the study is to investigate the theoretical limit of the predictability of indoor human mobility and its practical implications. The research team utilized an extensive dataset comprising 140 million motion sensor readings recorded in 117 smart homes. These homes were equipped with various sensors, such as passive infrared motion sensors, magnetic door sensors, and contact-based item sensors. The sensor data provided a detailed account of residents’ interactions with their environment, enabling a unique study of indoor movement patterns.
Methodological Approach
The methodology centered on the entropy rate and its relationship with predictability. The entropy rate is a measure of the average conditional entropy of a stochastic process and is crucial for estimating predictability. The researchers utilized theoretical concepts like Fano’s inequality and Jensen’s inequality, along with practical tools like data compression techniques (LZ77, LZMA, PPMD), to estimate the entropy rate from smart home data. These methods allowed them to empirically validate their theoretical analysis, offering a comprehensive understanding of indoor human mobility.
Key Findings and Insights
The findings of the study were revealing in several ways:
- Predictability Variation: The predictability of human mobility indoors varies significantly across different living arrangements. In homes with a single resident, the predictability of movement was found to be higher. In contrast, homes with multiple residents or pets showed lower predictability. This variation is attributed to the increased complexity in movement patterns in multi-resident homes and homes with pets.
- Impact on Smart Home Technology: These findings have crucial implications for smart home technology. The study suggests the need for more sophisticated models to predict human mobility accurately in various living environments. Understanding these limits of predictability can guide the design of more effective and personalized smart home systems.
- Entropy Rate and Predictability: The research demonstrated that the most powerful prediction model for indoor mobility could be characterized by the conditional probability distribution of the next sensor message based on the complete past history of resident trajectories. The maximum predictability was defined as the prediction accuracy of the next sensor message, averaged over the entire observed sequence. This approach provided a novel way to conceptualize and quantify indoor mobility predictability.
- Theoretical and Practical Implications: The theoretical analysis, coupled with empirical validation, offered a benchmark for evaluating and improving mobility models in smart homes. It also underscored the importance of considering different household compositions (like single or multiple residents, presence of pets) when designing predictive models for smart homes.
- Room for Improvement: The study concluded that there is potential for developing more representative models for indoor human mobility, especially in multi-resident settings. The results indicate that by incorporating additional information such as the time of day, the day of the week, and the length of resident stay at specific locations, the predictability of models can be significantly improved.
In summary, the paper presents a significant advancement in the field of smart home technology and human mobility prediction. By exploring the predictability of indoor human movement through a blend of theoretical and empirical approaches, the study offers vital insights for designing future smart home systems. These systems could be more responsive and adaptive to the needs of their residents, leading to enhanced living experiences and greater efficiencies in smart home operations. The research has the potential to influence not only smart home technology development but also broader applications in areas like elderly care, energy management, and security in residential settings.
The research appeared in the article “The Indoor Predictability of Human Mobility: Estimating Mobility with Smart Home Sensors” published in the IEEE Transactions on Emerging topics in Computing.