Theme 2

Data collection, security, and integration

  • 2.1: Robotics and Autonomous Systems for Data Collection

    Introduction

    There are many situations where data are required but potentially dangerous to collect. These include meta-stable, failed or at the verge of failure structure—both above and underground— whose state of integrity may be unknown. In such cases, a robot may provide the only option to safely carry appropriate sensors to collect data, map the structure, note damage, localise hazard and help assess integrity. Robots may be ground-based, able to carry heavy payloads; aerial, able to move quickly over difficult terrain; or submersible, for undersea inspection. Autonomous operation is essential where it is not possible to communicate with the robot because of interference from the environment.

    Innovation

    This subtheme will develop state-of-the-art algorithms for sensing, mapping and navigation that can be implemented on low-power GPUs. They will be based on robust position tracking (with Subtheme 1.2) and GPU based algorithm using Iterative Closest Point position tracking and Graph SLAM (Simultaneous Localisation and Mapping Algorithm) that can accurately generate a map of an unknown environment without the need for motion encoders. The approach will enable handling complex geometries and will create accurate 3D models of the environment. An adaptive goal reasoning method will be developed to perform task and role allocations to ground vehicles carrying heavy sensor packages and small UAV’s that can easily and quickly navigate confined spaces but can only carry small sensor packages. The multi-agent architecture used will include data fusion methods for combining maps from multiple vehicles.

    Outcome

    Autonomous systems capable of inspecting infrastructure including mines and other difficult environments; algorithms for sensing, mapping and navigation targeted for implementation on low-power GPUs, suitable for mobile platforms; novel multi-agent goal reasoning and task allocation methods.

  • 2.2: Big Data Management and Analytics in the Cloud

    Introduction

    To accurately capture the state of an infrastructure through sensing, the voluminous data gathered must be efficiently stored and managed to support critical analytics tasks to enable improved and enriched understanding of the infrastructure system status.

    Innovation

    This subtheme will develop advanced and innovative solutions for large- volume, dynamic and heterogeneous data management and analytics in the cloud, based on heterogeneous data modelling, storage and integration mechanisms, index construction, and maintenance strategies. Solutions developed will enable multiple continuous analytics tasks and monitoring of the analytics outputs in real-time with the dynamic updates of infrastructure data. Key developments will include: i) advanced and effective data models and storage engines to accommodate heterogeneous data captured by varying sensor devices, including structured, graph, time series, sequence, geospatial, image, and video data, ii) domain-specific data integration methods to efficiently link data from different data sources and disparate data stores to provide a unified view for data interpretation, iii) verified and empirically evaluated advanced algorithms for incremental index maintenance, iv) novel real-time distributed infrastructure data processing algorithms in the cloud, which can handle high-speed updates with real-time response, and v) targeted representative analytics, including aggregation computation, pattern mining and spatial-temporal search, and vi) space-, time- and communication-efficient techniques, along with rigorous analyses of the new techniques in terms of time complexity, space complexity, I/O cost, communication cost, scale dependency, and workload balance. These developments are critical for real-time response analysis and time-critical safety assessment of infrastructures performance (Subthemes 4.3, 5.2).

    Outcome

    A unified next-generation framework of big data management and real-time analytics in the cloud tackling large volume, dynamic updates and heterogeneity.

  • 2.3: Data security, robustness and reliability

    Introduction

    Remote sensing mechanisms are isolated embedded systems which collect information from sensors and store information locally until the information is transferred to a hub. Sensitive information is at risk of theft and alteration. A remote sensor system must therefore be secure against threats at different levels of abstractions. These include hardware security, side-channel analysis security, remote sensor system to hub security, software security and eavesdropping (snooping attacks) during wireless transmission. Industry and regulatory bodies require guarantees against threats and Standard Operating Procedures (SOP) for recovery when a threat occurs. The technical challenge is in creating and implementing robust countermeasures against attacks and rapid responses when attacks do occur.

    Innovation

    To address these challenges, this subtheme will develop countermeasures against electromagnetic, power and fault attacks on AES protocols; novel secure processor designs based on Advanced Risc Machines (ARM) and RISC-V processors; secure software systems using protocols such as Internet Protocol Security (IPSec), Secure Shell (SSH) and Transport Layer Security (TLS); and, robust solutions with above protocols and processors that can be adopted in embedded, remote systems. Protocols will be developed to identify the information leakage of authentication, the session key exchange (both PKI and Quantum key exchange will be studied) between the remote sensors and hub to initiate secure communication. Other safeguards developed will include: checkpoint recovery processors to reverse the effects of the data alterations; monitoring the processor against security breaches; and, countermeasures to mitigate timing vulnerabilities of the software by the use of secure kernels such as sel4.

    Outcome

    Robust, reliable circuits and system designs, which are capable of withstanding multitudes of side-channel attacks; remote sensor system security countermeasures to provide robust, remotely upgradable hardware and software countermeasures to mitigate attacks on wireless communication channels with minimal power overhead; protocols to protect the software stack in remote sensors from cyberattacks.

Chief Investigators

Claude Sammut

Chief Investigator
UNSW Sydney
View Bio

Tommy Chan

RIIS Hub Lead Chief Investigator
Queensland University of Technology
View Bio

Sridevan Parameswaran

Chief Investigator
UNSW Sydney
View Bio

Wenjie Zhang

Chief Investigator
UNSW Sydney
View Bio

Stuart Clark

Chief Investigator
UNSW Sydney
View Bio

Binghao Li

Chief Investigator
UNSW Sydney
View Bio

Jack Barton

Chief Investigator
UNSW Sydney
View Bio

Simit Raval

Chief Investigators
UNSW Sydney
View Bio

Johnson Xuesong Shen

Chief Investigator
UNSW Sydney
View Bio