Light Controlled Factory
Call Us: +44 (0)1225 385138

Research Theme 3

A ubiquitous, 7D measurement environment for the entire factory space

The real time positioning of parts and machines within an assembly factory requires multiple bodies to be accurately located (with accuracy from 500 to 10 μm) within the 3D space with 6 DOF data (x, y, z, roll, pitch, yaw) and data rates from a few tens to a thousand points per second, tracked over time (7D). Such an environment cannot be provided by any single system available today and will have to be designed and created by the synthesis of LVM and localised metrology systems. The WPs are as follows.

WP 3.1: Formal measurement network design models and algorithms (Maropoulos, Mullineux, Huntley,

Robson, Boehm). In the era of the Light Controlled Factory, industrial metrology must be treated as a production process and measurement process models will be created. The models will be; laser trackers and iGPS (Bath), structured light (Loughborough) and photogrammetry (UCL). The network design stages will include: (a) Understanding and codifying the factory’s needs in terms of process control and parts inspection

(b) Establishing the factory’s scale, spatial and environmental conditions.

(c) Identifying and selecting capable and networked measurement systems using the process models, on the basis of measurement uncertainty, scale, environmental control, instrument’s technological maturity, set-up and deployment times, and cost.

(d) Optimising the number, type and deployment sequence of systems to offer acceptable levels of data redundancy, optimise line of sight coverage and minimise uncertainty. This will involve the design of bundle adjustment and data fusion algorithms (with WPs 3.4 and 3.5).

WP 3.2. Novel metrology embedded in structures (Knight, Wadsworth, Maropoulos).

AIT structures often make direct measurement of the key assembly characteristics impossible. By embedding localised metrology, such as lasers and LVDTs within structures, it will be possible to utilise the assembly structure to extract direct measurement data for all key “jig to part” interfaces. The research will develop novel photonic techniques for absolute distance measurement using interferometry or coherence methods.

Large distances can be divided down with high precision using stable pulsed lasers and the remaining discrepancies ascertained by spectral interferometry of widely dispersed broadband pulses. Technical challenges include (i) the design and validation of the electromechanical interfaces of the laser and optics with the structural elements, (ii) the miniaturisation of the photonic components, and (iii) optical fibre distribution of a single laser to multiple metrology paths within a jig. Benefits include improved measurement accuracy and the elimination of line of sight constraints of external optical methods.

WP3.3 Modelling, measurement and mitigation of environmental influences in LVM (Robson, Boehm).

WP3.3 will address the accuracy and capability limitations of optical LVM systems by quantifying and modelling localised optical effects caused by variations in the environment. This challenge will be met by developing and deploying a network of low cost photogrammetric sensors able to coordinate many thousands of target locations, augmented by state of the art, high resolution sensors equipped with fisheye lenses similar to those recently calibrated by UCL for NASA. To further enhance the environmental sensitivity of the system, the project will develop a novel LED light source providing target illumination at narrow wavelength bands in the blue and near infra-red. Comparison between target sub-pixel image locations at each wavelength against a dense static target array will provide a second set of environmentally perturbed signals.

The system will provide a high degree of redundant information which can be inter-compared to detect environmental variations across the measurement volume.

WP 3.4. Algorithms and networks of sensors for network adjustment (Robson, Boehm, Maropoulos, Huntley).

Drawing upon the deliverables from WP3.3, the dense spatial and environmental information will be augmented in three ways:

  • through the inclusion of dual frequency photogrammetric models validated through laboratory experiments;
  • development of a new volume driven network adjustment approach whereby the measurement space is sub-divided into discrete blocks and image rays traced through the volumetric structure;
  • through the analysis of network adjustment error propagation patterns exposed as covariance and correlation between measurements from disparate instruments to common locations.

The result will be a new, volume driven network (bundle) adjustment approach.

WP 3.5. High-accuracy and high-density point clouds for surface metrology and pose estimation (Huntley).

Projected fringe systems can provide high-accuracy (~50 μm) high-density point clouds (~1 point mm-2), directly from surfaces of parts, machines and jigs used for surface measurement and the automated extraction of the position and orientation (pose) of parts. This is essential for the assembly process, for example identifying robot pick-up points, or part alignment in jigless assembly. Fast Hough transforms will be extended from the identification of geometric primitives, e.g., spheres and planes, to complete parts and more general curved surfaces in known relative orientations. Bayesian classification will be used for position and pose estimation, with refinement through a computationally efficient extension to the Iterative Closest Point algorithm. An additional aim, in collaboration with WP3.4, will be to develop rapidly-deployable artefacts, with bundle adjustment refinement of camera/projector models, to form a global coordinate system for multiple scanners over a volume of 4×3×3 m3 and an accuracy of <100 μm.

WP 3.6. Reconfigurable measurement networks in response to changing factory and product needs (Huntley, Robson, Maropoulos).

WP3.6 will research how flexible and reconfigurable networks of scanners could be established rapidly in response to new products and AIT cell layouts, in collaboration with WP3.1 and WP3.3 in terms of optimising network design and predicting measurement uncertainty. Another level of flexibility in the form of ‘smart’ 3D sensors is required to deal with the main systematic error sources, namely the response to; (i) shiny parts (e.g. freshly machined metal), which give specular ‘hot spots’, (ii) concave features that cause multiple reflection of the scattered light, (iii) moving components. New, adaptive algorithms to address each of these issues will be developed and validated.