Wireless InSite: Deterministic Ray-Tracing

• Wireless InSite is a deterministic ray-tracing tool that models site-specific wireless propagation with path-resolved outputs such as delay spread and power delay profiles.
• It integrates detailed 3D environment construction and material assignment to accurately capture multipath effects in urban, indoor, vehicular, and UAV scenarios.
• The tool supports advanced applications by leveraging physics-based models for reflections, diffractions, and transmissions, enabling precise simulation of complex channels.

Wireless InSite is a proprietary electromagnetic propagation software and commercially available tool for deterministic, site-specific wireless modeling. Across recent work, it is used as a 3D ray-tracing environment that converts explicit geometry, material properties, antenna configurations, and motion into path-resolved propagation outputs such as path loss, received power, time of arrival, direction of arrival/departure, delay spread, and complex impulse response (CIR). Its role is typically that of a “physics engine” for scenarios in which ensemble-averaged or purely geometric abstractions are insufficient, including dense urban outdoor-to-indoor channels, indoor floor plans, vehicular crossroads, UAV air-to-ground links, static mesh planning, and wireless digital twins (Macaia et al., 20 Mar 2026, Aram et al., 2024, Andrusenko et al., 22 May 2026).

1. Modeling paradigm and solver architecture

Wireless InSite is used in the cited literature as a deterministic ray-tracing platform rather than as a stochastic channel emulator. In the outdoor-to-indoor and indoor power-delay-profile comparison, it is run with the X3D shooting-and-bouncing-ray propagation engine, with GPU acceleration, multithreading, and exact path calculations; the resulting multipath is geometry-dependent and receiver-specific (Andrusenko et al., 22 May 2026). In urban digital-twin analysis, the X3D solver is characterized as a high-fidelity full 3D ray-tracing model that supports accurate predictions from about 100 MHz to 30 GHz and includes reflections, diffractions, transmissions, and atmospheric absorption (Aram et al., 2024).

The solver philosophy described around Wireless InSite is that of high-frequency electromagnetic approximation. One paper frames modern ray tracing in terms of shooting-and-bouncing rays based on geometrical optics, physical optics for scattered fields, and uniform theory of diffraction for diffraction (Aram et al., 2024). A separate dataset paper states that its communication data are based on ray tracing that follows geometrical optics and uniform theory of diffraction (Cheng et al., 2023). This places Wireless InSite within the class of site-specific path-building solvers that resolve direct, reflected, transmitted, diffracted, and, where configured, foliage- or weather-modified components.

The outputs emphasized in the literature are path-resolved rather than merely link-budget summaries. In indoor workflow studies, Wireless InSite is used to output dominant paths, path loss / received power, complex impulse response, delay spread, and path arrival times (Macaia et al., 20 Mar 2026). In the Washington, D.C., outdoor-to-indoor study, the complex impulse response outputs are singled out because they contain absolute received power, phase, and time of arrival for each propagation path, after which the paths are converted into power delay profiles (PDPs) using excess delay rather than absolute time of arrival (Andrusenko et al., 22 May 2026).

2. Geometry, materials, and environment construction

A defining feature of Wireless InSite usage is the explicit construction or import of 3D environment models. In an indoor propagation framework, measured building dimensions are first captured using a meter tape and a BOSCH GLM-40 pocket laser distance measurer, then modeled in SketchUp, exported as .kmz, and imported into Wireless InSite for ray tracing (Macaia et al., 20 Mar 2026). The modeling procedure described there includes drawing the 2D floor plan, extruding walls into 3D, creating openings, adding interior objects, assigning materials and textures, and then exporting the model for propagation analysis.

Other studies use different front ends but the same basic principle. In a wireless digital-twin study of downtown Gothenburg, the CAD workflow for Wireless InSite uses Blosm, a Blender add-on, to import terrain with about 30 m resolution, build 3D buildings from OpenStreetMap, place buildings on terrain, and support building heights, floor counts, and roof shapes; the premium version can also import forests and trees (Aram et al., 2024). In the M$^3$SC dataset, a vehicular urban crossroad scene built in AirSim is imported into Wireless InSite with some fine detail omitted, the color of each side of the building is modified, and the crosswalk is added, after which object dimensions and coordinates are aligned across AirSim, WaveFarer, and Wireless InSite (Cheng et al., 2023).

Material assignment is equally central. The Washington, D.C., study explicitly models terrain and building geometry and assigns electrical properties for asphalt, concrete, glass, and drywall (Andrusenko et al., 22 May 2026). In the Guam mesh-network planning study, the environment includes Wet Earth terrain, concrete buildings, and dense deciduous forest, “in leaf,” while the propagation run uses the Simple Attenuation foliage model (Andrusenko, 22 May 2026). In the UAV transfer-learning study, Wireless InSite maps include dielectric half-space materials such as concrete and wet earth and one-layer dielectric material such as brick (Sun et al., 20 Jan 2025).

These inputs are not merely cosmetic. The indoor framework verifies material sensitivity by simulating the same room with wooden furniture and metal-coated furniture; the reported effect is that wooden furniture leads to stronger absorption and fewer dominant reflected rays, whereas metallic furniture produces stronger reflections and more visible multipath components (Macaia et al., 20 Mar 2026). A plausible implication is that Wireless InSite is most informative when the scene description is sufficiently detailed to distinguish not only walls and openings but also materially distinct indoor or urban objects.

3. Channel representations, metrics, and post-processing

The most common post-processing path around Wireless InSite begins with the CIR. In the indoor workflow paper, CIR or arrival-time data exported from Wireless InSite are processed in MATLAB to generate a PDP, and the PDP can then be used to build a tapped-delay-line channel model for downstream link-level studies such as BER simulation (Macaia et al., 20 Mar 2026). In M$^3$SC, the channel is written as $h(\varepsilon, t, \tau)$, with a spatial-temporal varying transfer function

$$\phi(\varepsilon, t, f)=\int h(\varepsilon, t, \tau) e^{-j 2 \pi \tau f}\,\mathrm{d}\tau,$$

from which the space cross-correlation function, time auto-correlation function, and Doppler power spectrum density are derived by setting selected offsets to zero and Fourier transforming the TACF with respect to $\Delta t$ (Cheng et al., 2023).

In the outdoor-to-indoor versus 3GPP comparison, Wireless InSite PDPs are power-normalized to enable comparison of relative multipath delay structure. The procedure is explicit: each Wireless InSite PDP is normalized to its peak tap, a $-30$ dB threshold relative to the peak tap is applied, and the PDPs are linearized, normalized, and interpolated before Kullback-Leibler divergence is computed (Andrusenko et al., 22 May 2026). The metrics used there are RMS delay spread, mean excess delay, effective maximum delay, and KL divergence. The standard mean excess delay is

$\bar{\tau} = \frac{\sum_i P_i \tau_i}{\sum_i P_i},$

and the RMS delay spread is

$$\sigma_\tau = \sqrt{\frac{\sum_i P_i (\tau_i - \bar{\tau})^2}{\sum_i P_i}}.$$

The effective maximum delay is the delay of the last significant multipath component retained above the $-30$ dB threshold, and the KL divergence is interpreted in bits, with $D_{KL}=0$ denoting a perfect match (Andrusenko et al., 22 May 2026).

Wireless InSite outputs are also used for sanity checking and physical consistency verification. In the NIT Durgapur indoor study, the first-arriving LOS path is checked against geometric distance using

$^3$0

with $^3$1, $^3$2, and $^3$3, producing an approximately matching $^3$4 (Macaia et al., 20 Mar 2026). This use of first-arrival consistency as a scaling check recurs conceptually in many geometry-based ray-tracing workflows even when it is not written out explicitly.

4. Research uses across indoor, outdoor, vehicular, and aerial systems

Wireless InSite is used for markedly different research tasks, but the common pattern is the extraction of site-specific electromagnetic structure from an explicit scene model. In dense urban outdoor-to-indoor and indoor channels, it serves as the deterministic reference for comparing site-specific PDPs against 3GPP TR 38.901 TDL-A/TDL-B/TDL-C NLOS models in a 1 km $^3$5 1 km Dupont Circle area in Washington, D.C., with 666 indoor receivers on the first floor of a representative eight-story building at 3.5 GHz (Andrusenko et al., 22 May 2026). In that setting, the emphasis is on late-arriving energy, irregular spikes, clustered multipath caused by geometry, and receiver-specific PDP structure.

In indoor deployment planning, Wireless InSite is the core of an end-to-end pipeline in which a measured floor of the Electronics and Communication Engineering department of NIT Durgapur is modeled and analyzed for coverage-hole prediction (Macaia et al., 20 Mar 2026). The main application example places a single isotropic transmitter at the eastern side of the corridor and 16 receivers across rooms and labs, using a sinusoidal signal at 1 GHz, 10 kHz bandwidth, a maximum number of rays of 25, and antenna height of 2 m. The reported result is that the transmitter location is not optimal, with many coverage holes, particularly poor signal availability in the faculty rooms, and better signal quality in the corridor than in some room interiors (Macaia et al., 20 Mar 2026).

In integrated sensing and communication research, Wireless InSite supplies the electromagnetic-space data of the M$^3$6SC dataset. It generates ray-tracing-based CIRs for V2V and V2I links in a vehicular urban crossroad scenario containing 11 base stations, 12 vehicles, 6 pedestrians, trees, and buildings, under sunny, rainy, and snowy conditions, at 5.9 GHz with 20 MHz bandwidth and at 28 GHz with 2 GHz bandwidth (Cheng et al., 2023). The dataset contains 1500 snapshots, and Wireless InSite contributes 108,000 CIR matrices overall (Cheng et al., 2023).

In UAV communications and learning-based control, Wireless InSite underpins both channel characterization and policy training. For mmWave air-to-ground characterization, it is used to emulate a fixed ground transmitter at 2 m and a UAV moving over roughly 2 km in urban, suburban, rural, and over-sea environments at 28 GHz and 60 GHz, extracting RSS and RMS delay spread (Khawaja et al., 2017). In transfer-learning-based UAV trajectory training, it produces high-fidelity SINR / outage-rate maps over Ottawa and Rosslyn from real-world urban layouts, using detailed 3D building models, material-aware propagation, a dense 5 m receiver grid at UAV altitude, and four terrestrial base stations with half-wave dipole antennas transmitting at 30 dBm (Sun et al., 20 Jan 2025).

Wireless InSite is also used as the front end of planning frameworks whose later stages are not themselves propagation simulations. In the Guam static mesh-network study, it generates a full $^3$7 path-loss matrix at 925 MHz for 155 fixed infrastructure sites, after which the matrix is symmetrized, converted into a similarity graph, and fed into spectral embedding, balanced k-means clustering, and gateway selection (Andrusenko, 22 May 2026). In a digital-twin study of downtown Gothenburg, it is used to produce route-based RSS profiles and coverage heat maps for alternative building layouts, thereby supporting the claim that small geometric changes in urban layouts can strongly affect coverage (Aram et al., 2024).

5. Comparative position relative to stochastic, lightweight, and learned models

The most direct comparison in the cited literature is between Wireless InSite and ensemble-averaged channel models. In the Washington, D.C., study, 3GPP TDL models generally exhibit longer delay spreads and longer tails than the site-specific Wireless InSite PDPs; for outdoor-to-indoor cases, RMS delay spread, mean excess delay, and maximum excess delay are always longer in the 3GPP models than in the Wireless InSite PDPs (Andrusenko et al., 22 May 2026). The same study states that TDL models may omit late-arriving deterministic energy and isolated strong spikes because they use fixed tap delays and powers and are not mapped uniquely to a specific environment geometry (Andrusenko et al., 22 May 2026). The practical distinction is therefore not simply deterministic versus stochastic, but geometry-preserving versus ensemble-averaged structure.

A second comparison concerns lighter-weight or more automated ray tracers. In the Gothenburg digital-twin work, Matlab’s Communications Toolbox ray tracer is easier to connect to OpenStreetMap-based scene creation, but the paper highlights that Matlab does not extract tree/forest information and that custom geometry augmentation is more difficult once the coordinate system is fixed (Aram et al., 2024). Along a 222-receiver L-shaped route, Matlab predicts RSS values that are on average 15–20 dB higher than Wireless InSite in the first part of the route; the paper suggests that Wireless InSite’s inclusion of the tree line and Matlab’s lower diffraction order are the main reasons (Aram et al., 2024).

A third line of comparison places Wireless InSite against alternative site-specific engines or surrogates. The DCEM paper on smooth-surface diffraction situates its own simulator in the same family as commercial tools such as Wireless InSite or WinProp, but argues that additional UTD treatment of smooth convex diffraction is needed for more accurate shadow-boundary prediction in indoor scenes (Wang et al., 2023). GeNeRT, by contrast, uses Wireless InSite in two roles: as a label/data-generation tool for training and evaluation and as a runtime baseline for a physics-informed neural ray-tracing surrogate (Bian et al., 23 Jun 2025). The paper states that when there is only one transmitter, runtime of Wireless InSite and GeNeRT is comparable, but as the number of transmitters increases, GeNeRT becomes increasingly faster than Wireless InSite (Bian et al., 23 Jun 2025).

Wireless InSite also appears as a baseline in digital-twin systems designed to be more open or programmable. A 3D Wi-Fi signal measurement system built from LiDAR-scanned geometry, semantic segmentation, ITU-R material annotation, and GPU-accelerated ray tracing is explicitly described as a digital-twin, ray-tracing-based alternative to Wireless InSite and similar commercial tools (Wang et al., 12 Feb 2026). This suggests a broader methodological trend: Wireless InSite functions both as a production-grade deterministic solver and as a reference point against which newer digital-twin or learned simulators are evaluated.

6. Limitations, assumptions, and appropriate use

The limitations attributed to Wireless InSite in the cited work are mostly limitations of deterministic ray-tracing workflows rather than software defects in isolation. First, results depend on the fidelity of the scene model. In M$^3$8SC, the AirSim scene is imported into Wireless InSite with some fine details omitted, so the electromagnetic and physical spaces are aligned but not perfectly identical in all detail (Cheng et al., 2023). In the Washington, D.C., indoor study, architectural blueprints were unavailable, and a representative floorplan was replicated across all eight floors even though the analysis focuses on the first-floor receiver grid (Andrusenko et al., 22 May 2026).

Second, the accuracy of site-specific prediction depends on propagation assumptions and configured mechanisms. The M$^3$9SC paper notes that rainy and snowy days are imitated in Wireless InSite by introducing rainy and snowy models with shapes set as raindrops and snowflakes, while weather condition simulation in WaveFarer is currently underway (Cheng et al., 2023). The same paper also notes that dynamic vehicular motion is discretized into 1500 snapshots rather than modeled continuously (Cheng et al., 2023). A plausible implication is that time variation is represented as a sequence of static solves whose resolution is bounded by snapshot density.

Third, Wireless InSite is not interchangeable with standardized stochastic models. The outdoor-to-indoor PDP comparison concludes that 3GPP TDL models are suitable for large-scale system evaluation, whereas deterministic or hybrid approaches are more appropriate for site-specific physical-layer design, especially when multipath delay structure matters, cyclic prefix sizing is important, intersymbol interference sensitivity must be assessed, or late-arriving energy and irregular spikes could affect performance (Andrusenko et al., 22 May 2026). This is less a critique of either model class than a domain-of-validity statement.

A common misconception is that similar delay-spread statistics imply similar channels. The Washington, D.C., study explicitly cautions that even if the overall delay spread seems similar, the actual PDP shape may be very different because Wireless InSite preserves late-arriving energy, irregular spikes, clustered multipath caused by geometry, and receiver-specific structure tied to walls, windows, reflections, and building layout (Andrusenko et al., 22 May 2026). Conversely, another misconception is that deterministic ray tracing automatically removes the need for calibration or verification. The indoor workflow paper performs delay-distance consistency checks and material-change sanity checks (Macaia et al., 20 Mar 2026), and the digital-twin Wi-Fi system that positions itself against Wireless InSite still requires calibration to align absolute power with measurements despite achieving high spatial correlation (Wang et al., 12 Feb 2026).

Taken together, the literature positions Wireless InSite as a deterministic, geometry-aware propagation engine whose distinctive value lies in preserving site-specific multipath structure. It is most appropriate when the environment geometry and fine channel structure matter directly, and less central when only standardized ensemble-average behavior is required (Andrusenko et al., 22 May 2026).