Chapter 9

Work plan

Work packages, tasks, schedule, milestones and decision gates.

The project runs for 52 weeks with a team of four (Chapter 10). The work is divided into seven work packages: a continuous management package (WP0) and six technical packages. Each technical package ends with a measurable milestone. Four decision gates define what happens if a milestone is missed.

9.1Structure and logic

The order of the work follows the dependencies between measurements. The receive chain must work (WP1) before the static rig can be calibrated against theory (WP2). The full-tensor dataset (WP4) needs a calibrated rig and, for tracked captures, the positioning platform (WP3). The radar (WP5) is developed in parallel because it has no dependency on the induction hardware. Fusion, field trials and release (WP6) close the project.

Table 9.1.Work packages.

WP	Title	Lead	Weeks	Objective
WP0	Project management	PI	1–52	Planning, reviews, procurement, documentation, reporting
WP1	Bring-up and receive chain	HW	1–8	Working codec, coils and preamplifiers with measured characteristics
WP2	Static induction rig and calibration	PI	6–18	Measured sphere spectra agreeing with the exact model
WP3	Positioning and timing platform	SW	10–26	Pose of the coil at every block to 2 cm
WP4	Full tensor, dataset and classification	DS	18–36	Labelled dataset and a classifier validated on unseen objects
WP5	Radar subsystem	HW	22–40	SFCW radar imaging metal targets at 35 cm
WP6	Fusion, field trials and release	PI	36–52	Fused system tested in the field; public release

9.2Schedule

Figure 9.1.Schedule at task level. Bars are coloured by the role that leads the task; diamonds mark milestones; weeks 1 to 52.

9.3Tasks

Each task has one responsible owner (see the role codes in Chapter 10), a time window and a defined result. The acceptance criterion states how completion is checked.

Table 9.2.Tasks with owner, supporting role, time window and acceptance criterion.

ID	Task	Lead	Support	Weeks	Acceptance criterion / result
WP0 · Project management (lead PI, weeks 1–52)
T0.1	Kick-off; project handbook, repository and data conventions	PI	all	1–3	Handbook D0.1 agreed by all four team members
T0.2	Procurement and budget control	PI	SW	1–52	Orders placed per procurement schedule; monthly spending report
T0.3	Weekly meetings, monthly reviews, risk and decision logs	PI	all	1–52	Minutes and logs up to date at every monthly review
T0.4	Maintenance of this report; quarterly reports	PI	DS	1–52	Report revised monthly; quarterly reports on time
WP1 · Bring-up and receive chain (lead HW, weeks 1–8)
T1.1	Bench set-up: instruments, supplies, single-point grounding	HW	–	1–2	Bench ready; ground scheme documented
T1.2	Codec bring-up on Teensy 4.1 and Raspberry Pi; host decision	SW	HW	1–4	All 14 channels verified in loopback; no lost samples in 1 h; host chosen
T1.3	Codec characterisation: full scale, input noise, coupling corner, 96 kHz test	HW	SW	3–5	Measured values replace datasheet values in Chapter 5 (D1.1)
T1.4	Wind and measure transmit and receive coils	HW	–	2–5	Self-resonance at least 400 kHz; inductance within 10% of 4.58 mH
T1.5	Preamplifier and calibration-injector PCB: design, fabrication, test	HW	SW	3–8	Input noise at most 3 nV/√Hz; channel gains matched within 1% (D1.2)
T1.6	Acquisition software: tone generation, block demodulation, logging	SW	DS	2–7	12-tone loopback stable to 0.1% amplitude and 0.1° phase over 10 min
WP2 · Static induction rig and calibration (lead PI, weeks 6–18)
T2.1	Transmit driver: class-D amplifier, series resistor, current sense, soft ramps	HW	–	6–9	1 A per coil without clipping at crest factor 3.1
T2.2	Coil geometry and nulling fixture	HW	PI	8–13	Direct coupling at most 10⁻³ at all 12 tones (D2.2)
T2.3	Forward model: exact sphere, dipole model, coil geometry	PI	–	6–10	Model reproduces the three analytic limits and published examples
T2.4	Calibration measurements with four sphere types at 10, 20 and 30 cm	PI	DS	13–16	Complete data set including repeat measurements
T2.5	Comparison with the model; systematic error budget	PI	DS	15–18	M1 criterion evaluated; calibration report D2.1
T2.6	Ground survey at two sites; ground-response factor	DS	HW	14–18	Background spectra and measured ground factor for both sites
WP3 · Positioning and timing platform (lead SW, weeks 10–26)
T3.1	Sensor pole: non-metallic structure, mounts, lever arm	HW	–	10–13	Rotation test: computed coil position stays within a 2 cm circle
T3.2	RTK base and rover, correction link, availability survey	SW	–	10–14	Fixed solution for at least 95% of the time at the test sites
T3.3	PPS timing into codec and camera	SW	–	12–16	Timing error at most one sample against GNSS time tags
T3.4	IMU and magnetometer integration; interference test with transmitter on	SW	HW	13–16	Heading change at most 1° when the transmitter switches on
T3.5	Camera visual-inertial odometry with height scaling	DS	SW	14–20	Drift at most 1% of distance on a 20 m test track
T3.6	Extended Kalman filter: RTK, IMU, camera, magnetometer, height	SW	DS	16–24	Continuous pose at block rate (D3.1)
T3.7	Tracked-pass validation: peak width against depth, repeatability	PI	DS	22–26	M2 criterion evaluated; validation report D3.2
WP4 · Full tensor, dataset and classification (lead DS, weeks 18–36)
T4.1	Three-axis transmit head with interleaved tone sets	HW	SW	18–24	Cross-talk between transmitters at most -60 dB after FFT separation
T4.2	Tensor inversion software	PI	DS	20–28	Simulated tensors recovered within 5% at 30 dB SNR (D4.2)
T4.3	Simulator: spheres, spheroids, finite-element shapes; noise and soil	DS	PI	20–30	10,000 simulated captures; agreement with calibration data
T4.4	Dataset collection: at least 30 objects, 500 captures, 2 soils	DS	all	24–34	Dataset v1 with metadata and data sheet (D4.1)
T4.5	Classifiers: gradient boosting baseline, 1D-CNN comparison	DS	PI	28–35	Cross-validated accuracy with confidence intervals
T4.6	Blind test on hidden objects	DS	HW	34–36	M3 criterion evaluated; blind-test report D4.3
WP5 · Radar subsystem (lead HW, weeks 22–40)
T5.1	Pluto set-up: firmware profile, two-channel mode, coupler loopback	HW	SW	22–25	Reference-corrected phase stable within 2° over 100 retunes
T5.2	Vivaldi antennas: simulation, fabrication, measurement	HW	–	22–30	S11 at most -10 dB from 0.5 to 3 GHz
T5.3	Sweep software, background subtraction, range profiles	SW	PI	24–32	Metal plate in air at the correct range within 2 cm
T5.4	Amplifiers, shielding and emission check	HW	–	30–33	Emission within applicable limits; radar hardware D5.1
T5.5	Sandbox radar tests along tracked passes	PI	DS	32–40	M4 criterion evaluated; radar report D5.2
WP6 · Fusion, field trials and release (lead PI, weeks 36–52)
T6.1	Fusion of induction and radar detections	DS	PI	36–44	Fused false-alarm rate lower than either sensor alone
T6.2	Live map and operator interface	SW	–	38–46	Map updated within 1 s of acquisition during a pass
T6.3	Field trials at two sites with seeded targets	PI	all	42–48	Trial report with ground truth by excavation (D6.1)
T6.4	Documentation of hardware, software and dataset	SW	all	40–52	An outside reader rebuilds the receive chain from the documents
T6.5	Review of patents and non-English literature	PI	–	40–44	Review added to Chapter 2
T6.6	Paper preparation and submission	PI	all	44–52	Paper submitted (D6.2)
T6.7	Public release of designs, code, dataset and report	SW	all	50–52	Release published (D6.3); M5

9.4Milestones

Table 9.3.Milestones and acceptance criteria.

ID	Week	Milestone	Criterion	Verification	Owner
M1	18	Calibration agrees with theory	Spectra of the four calibration spheres agree with the exact model within ±10% in amplitude and ±5° in phase at all 12 tones, at 10 and 20 cm	Calibration report D2.1, reproduced independently from raw data by DS	PI
M2	26	Positioning validated	Ten passes over a marked target from different headings give positions within 2 cm (1σ); depth from peak width within ±1.5 cm at 20 cm	Validation report D3.2	SW
M3	36	Classifier validated on unseen objects	Blind test: ferrous / non-ferrous accuracy at least 95%, shape class at least 85%, size within ±25% for at least 80% of objects	Blind-test report D4.3; objects hidden by HW	DS
M4	40	Radar imaging	A metal can of about 10 cm buried at 35 cm in dry sand is detected and located; depth error at most 3 cm; echo polarity identified as metal	Radar report D5.2	HW
M5	52	Fused demonstration and release	Live fused survey of a seeded 5 m × 5 m plot; designs, software, dataset and report published	Public repository; field-trial report D6.1	PI

9.5Decision gates

At each gate the team reviews the milestone evidence and chooses one of the prepared options. Decisions and their reasons are recorded in the decision log.

Table 9.4.Decision gates.

Gate	Week	Condition	If met	If not met
G1	18	M1 met	Full-tensor work (WP4) starts as planned	Up to 4 weeks of diagnosis of model or hardware; WP4 start moves accordingly, WP3 continues. If still not met in week 22, the dataset uses empirically calibrated spectra and the discrepancy is reported.
G2	26	M2 met	Tracked dataset collection with the hand-held pole	A wheeled cart with odometry replaces the hand-held pole; RTK used for start and end points of each track.
G3	36	M3 met	WP6 proceeds as planned	Data collection extended by 4 weeks; the operator interface (T6.2) is reduced to a minimum.
G4	40	M4 met	Radar included in fusion and field trials	Induction-only results are published; the radar is reported as a partial result and future work.

9.6Critical path and dependencies

The critical path runs T1.2 → T1.5 → T2.2 → T2.4 → M1 → T4.1 → T4.4 → T4.5 → M3 → T6.1 → T6.3 → M5. Its slack is small: a delay in nulling the coils (T2.2) or in the calibration measurements (T2.4) moves the dataset collection and all later work. The positioning platform (WP3) and the radar (WP5) have about four weeks of slack each. The main external dependency is procurement of the RTK receivers and printed circuit boards, which is started in week 1 (T0.2).

Table 9.5.Procurement schedule.

Item	Order in week	Needed in week	Remark
Teensy 4.1, op-amps, passives, connectors	1	2	Local or fast-shipping distributors
Magnet wire, foil, coil formers	1	2
Preamplifier PCB, first run	4	6	Design completed in T1.5
Class-D amplifier modules, sense resistors	3	6
RTK receivers, antennas, radio link	1	10	Longest lead time, import
IMU, magnetometer, camera, ranger	6	12
Calibration spheres, test objects, sandbox	6	12
Vivaldi antenna PCBs, RF cables, coupler	20	24	After simulation in T5.2
Second PCB run (revised preamp, tensor head)	16	20

9.7Deliverables

Table 9.6.Deliverables.

ID	Deliverable	Type	Due week	Owner
D0.1	Project handbook: lab-log template, data conventions, safety rules	Document	3	PI
D1.1	Codec bring-up and characterisation report	Report	5	SW
D1.2	Preamplifier and calibration-injector PCB, revision A	Hardware	8	HW
D2.2	Static induction rig (1 TX + 3 RX), documented	Hardware	13	HW
D2.1	Calibration report: measured sphere spectra against the model	Report	18	PI
D3.1	Positioning and timing platform with fusion software	Hardware, software	24	SW
D3.2	Tracked-pass validation report	Report	26	PI
D4.2	Tensor inversion software	Software	28	PI
D5.1	Radar hardware: antennas, amplifiers, coupler	Hardware	33	HW
D4.1	Labelled dataset, version 1	Data	34	DS
D4.3	Classifier and blind-test report	Report	36	DS
D5.2	Radar software and test report	Software, report	40	PI
D6.1	Field-trial report of the fused system	Report	48	DS
D6.2	Paper submitted to a journal or conference	Publication	52	PI
D6.3	Public release: designs, code, dataset, report	Release	52	SW

9.8Reporting and reviews

Weekly: one-hour team meeting (progress against tasks, blockers, next week's measurements). Short written notes are added to the lab log.
Every two weeks: demonstration of a working result on real hardware or real data.
Monthly: review of schedule, budget and risk register; update of this report.
At each milestone: milestone review with a written report, followed by the decision gate where one applies.
Quarterly: progress report to the funder with spending against budget.