Search

Letting robots emulate human behavior has always posed a challenge, particularly in scenarios involving multiple robots. In this paper, we presented a framework aimed at achieving multi-agent reinforcement learning for robot control in construction tasks. The construction industry often necessitates complex interactions and coordination among multiple robots, demanding a solution that enables effective collaboration and efficient task execution. Our proposed framework leverages the principles of proximal policy optimization and developed a multi-agent version to enable the robots to acquire sophisticated control policies. We evaluated the effectiveness of our framework by learning four different collaborative tasks in the construction environments. The results demonstrated the capability of our approach in enabling multiple robots to learn and adapt their behaviors in complex construction tasks while effectively preventing collisions. Results also revealed the potential of combining and exploring the advantages of reinforcement learning algorithms and inverse kinematics. The findings from this research contributed to the advancement of multi-agent reinforcement learning in the domain of construction robotics. By enabling robots to behave like human counterparts and collaborate effectively, we pave the way for more efficient, flexible, and intelligent construction processes., Comment: 19 pages

Construction robots are challenging the traditional paradigm of labor intensive and repetitive construction tasks. Present concerns regarding construction robots are focused on their abilities in performing complex tasks consisting of several subtasks and their adaptability to work in unstructured and dynamic construction environments. Imitation learning (IL) has shown advantages in training a robot to imitate expert actions in complex tasks and the policy thereafter generated by reinforcement learning (RL) is more adaptive in comparison with pre-programmed robots. In this paper, we proposed a framework composed of two modules for imitation learning of construction robots. The first module provides an intuitive expert demonstration collection Virtual Reality (VR) platform where a robot will automatically follow the position, rotation, and actions of the expert's hand in real-time, instead of requiring an expert to control the robot via controllers. The second module provides a template for imitation learning using observations and actions recorded in the first module. In the second module, Behavior Cloning (BC) is utilized for pre-training, Generative Adversarial Imitation Learning (GAIL) and Proximal Policy Optimization (PPO) are combined to achieve a trade-off between the strength of imitation vs. exploration. Results show that imitation learning, especially when combined with PPO, could significantly accelerate training in limited training steps and improve policy performance., Comment: 22 pages, 8 figures

Continuous construction and rehabilitation in urban settings have unavoidable impacts on arrival times of first responders to emergency locations. Current research efforts on emergency response assessments focus on case studies, where specific periods (e.g., super storm Sandy) of emergency response times are analyzed. Simulation based studies that aim to evaluate response times in relation to various constraints/fleet sizes also exist. However, they do not analyze how specific changes (e.g., new and ongoing construction projects) in urban settings impact emergency response times of first responders. This paper aims to fill the gap and proposes a novel approach to predict the expected emergency response time for a given location using the fabric of zones regarding construction activities. This approach relies on historical records of emergency response and construction permits issued by city agencies. The approach first defines the signature of a zone (by zip codes) for construction activities based on the distribution of historical construction work types permitted in that zone over time. Then, zones that share similar signatures are clustered to find if there exists a relationship between construction signatures and emergency response times. Next, supervised learning algorithms are deployed to predict the average emergency response times for each cluster. The approach was tested using New York City's construction permit and emergency response records, and can be easily replicated for other cities with similar public datasets. This study serves as the first step towards quantitatively understanding construction projects' impact on a quality of life (QoL) indicator (specifically emergency response times) in urban settings.

The impact of built environment on the human restorativeness has long been argued; however, the interrelations between neuroscience and the built environment, and the degree to which the built environment contributes to increased human restorativeness has not been completely understood yet. Understanding the interrelations between neuroscience and the built environment is critical as 90% of time in a typical day is spent indoors and architectural features impact the productivity, health and comfort of occupants. The goal of this study is to bring a structured understanding of architecture and neuroscience interactions in designed facilities and quantification of the impact of design on human experience. The authors first built two virtual environments (i.e., restorative and non-restorative) using the architectural designs features related to human restorativeness identified by previous research efforts. Next, user experiments were conducted in the two built virtual environments including 22 people. The subjects were asked to conduct navigational tasks while their bodily responses recorded by body area sensors (e.g., EEG, GSR, and Eye-tracking). The result showed that human responses in restorative and non-restorative environment had statistically significant difference. This study serves as the first step of understanding human responses in the virtual environment, and designing spaces that maximize human experience.

According to the World Bank, more than half of the world's population now lives in cities, creating burdens on the degraded city infrastructures and driving up the demand for new ones. Construction sites are abundant in already dense cities and have unavoidable impacts on surrounding environments and residents. However, such impacts were rarely quantified and made available to construction teams and local agencies to inform their planning decisions. A challenge in achieving this was the lack of availability of data that can provide insights about how urban residents respond to changes in their environment due to construction projects. Wider availability of data from city agencies nowadays provides opportunities for having such analysis possible. This paper provides the details of a generic data-driven approach that enables the analysis of impact of construction projects on quality of life in urban settings through the quantification of change on widely accepted quality of life indicators in cities. This paper also evaluated the approach using data from publicly construction projects' information and open city data portals from New York City. Historical 311 Service Requests along with 27 road reconstruction projects were used as testbeds. The results showed that 61% of the projects analyzed in this testbed experienced higher 311 requests after the commencement of construction, with main complaints of 'noise', 'air quality', and 'sewer' at the beginning of construction, and 'sanitation' and 'waste' towards the end. Prediction models, built using regression machine learning algorithms, achieved an R-Squared value of 0.67. The approach is capable of providing insights for government agencies and construction companies to take proactive actions based on expected complaint types through different phases of construction.

Protecting the passengers' safety and increasing ridership are two never ending pursuits of public transit agencies. One of the proposed methods to achieve both goals for subway service is to implement real time train arriving countdown clocks in subway stations. Metropolitan Transportation Authority (MTA) of New York City (NYC) chose to install such countdown clocks in their stations starting from 2007 on a selection of subway lines. Due to the recent development of Bluetooth Beacon technology, the MTA could now install countdown clocks and train trackers in a non intrusive manner with much faster speed. As a result, the MTA is aiming to install countdown clocks in every subway station on every line. However, with such an aggressive plan, the impact of countdown clocks on subway ridership has not been fully studied. This paper proposes using Panel Regression methods, specifically, Random Effect (RE) model and Fixed Effect (FE) model to quantify the impact of countdown clocks on subway ridership. Machine Learning methods, namely Random Forest (RF) with AdaBoost and Decision Tree (DT) Regression, are also used as alternative data driven approaches for the FE and RE model. The results show that for the G line service, which runs between Brooklyn and Queens, the introduction of countdown clocks could increase weekly ridership by about 1783 per station. The study also found that the machine learning methods provide better accuracy in predicting the ridership than RE and FE models.