SBIR-STTR Award

SSCI and Olin College of Engineering propose to develop, integrate and test an innovative IMPACT-HCP system (Integrated Mission Planning and Autonomous Control Technology for Heterogeneous Collaborating Platforms). The main features of the proposed system include several innovative behaviors for UAV/UGV collaboration for efficient mission execution. Proposed behaviors include: (i) Collaborative Landing Zone (LZ) Determination; (ii) Collaborative Team Navigation; (iii) Collaborative Mapping; (iv) Collaborative Landing; and (v) Coordinated Arrival. IMPACT-HCP learns about an uncertain environment in real time and performs decentralized task allocation using a Multi-step Allocation Algorithm (MA2). The integrated system will be implemented using a scalable software and communication architecture. IMPACT-HCP will be tested through hardware experiments at Olin College of Engineering at their 150-acre test range. The experiments will involve a USV/UAV/UGV team and collaborative behaviors developed under the IMPACT-HCP framework will be tested on a realistic mission scenario. The following capabilities will be demonstrated: image processing for search, detection and classification; dynamic path planning; sensor fusion for collaborative navigation and mapping; coordinated arrival; and coordinated landing of a UAV on a UGV and a USV. Phase II will focus on the full system prototype development and its testing under realistic operating conditions.

Benefit:
Autonomous collaboration between heterogeneous platforms has a great potential to substantially improve the performance of military missions and commercial applications. Search and rescue missions would benefit from IMPACT-HCP where the UGVs, guided by UAVs, would travel on unobstructed paths and be able to accurately detect and geolocate victims, and convey that information to the rescue teams. Integrated collaborating UAV/UGV teams will also find applications in precision agriculture. The ground and aerial measurements collected by the system can be used for estimating Nitrogen levels across a farm field, and these estimates can in turn guide fertilizer application. The capability to apply the right amount of fertilizer at the right time can improve productivity and drastically reduce fertilizer usage which is desirable from an environmental and economic standpoint. UAV/UGV teams can also be effectively used for land mapping and terrain classification.

Keywords:
Unmanned Aerial Vehicles, Unmanned Aerial Vehicles, autonomous control, Unmanned Ground Vehicles, Unmanned Surface Vehicles, Collaborative Behaviors

Natural disasters and battle can cause immense damage to visible structures such as buildings, roadways, power-lines, but also to underground gas, power, and water infrastructure. This damage can cause significant safety hazards and be catastrophic to the way-of-life for locals while also hindering disaster response times, assessments, and repair. Disaster response efforts typically require heavily trained teams to manually operate complex equipment to assess and localize damage to infrastructure. After damage assessment is performed, there is typically another manual effort to plan the reconstruction of the infrastructure. The goal of our proposed “Multi-Layer Understanding of urban Infrastructure and Response” (MUIR) approach is to automate and expedite these and similar assessment and response efforts. This shall be accomplished by enabling lightly trained teams to rapidly deploy one or more pre-configured toolboxes with pre-packaged sensors, platforms, and algorithms to quickly map infrastructure, assess damage, and virtually reconstruct the infrastructure while accounting for uncertainty and possibly inaccurate priors (e.g. subject matter expertise, utility maps). Our proposed solutions will allow first responders faster situational awareness for timely and impactful positioning and planning of resources, as well as for proper deployment of assistance, while also assessing the potential safety risks/threats in the region. Our overall technical solution includes (1) a configurable toolbox of heterogeneous platforms/sensors, (2) deployment of assets, (3) automated scene understanding algorithms (e.g. object detection, scene segmentation, contextual structural constraints) (4) damage assessment (e.g. deep network detectors and anomaly detection) to locate damage, with (5) processing on the platforms and ground station to (6) virtually reconstruct the infrastructure (7) optionally incorporate human adjudication, all while accounting for prior knowledge (8). This is all in the presence of incomplete, inaccessible, and/or dynamic observations that can change due to natural factors such as wind, rain, or snow. Infrastructure reconstruction will allow a more optimal allocation of both manned and unmanned vehicles operating in a highly dynamic and risky environment. MUIR proposes a solution capable of supporting various sensors and platforms while accounting for scalability.